CVE-2016-3304 : Detail

The EPSS model produces a probability score between 0 and 1 (0 and 100%). The higher the score, the greater the probability that a vulnerability will be exploited.

The percentile is used to rank CVE according to their EPSS score. For example, a CVE in the 95th percentile according to its EPSS score is more likely to be exploited than 95% of other CVE. Thus, the percentile is used to compare the EPSS score of a CVE with that of other CVE.

Source: https://bugs.chromium.org/p/project-zero/issues/detail?id=828 The Microsoft GDI+ implementation of the EMF format supports records corresponding to the ExtTextOutA() and PolyTextOutA() API functions. Both CEmfPlusEnumState::ExtTextOutA and CEmfPlusEnumState::PolyTextOutA handlers suffer from a security vulnerability in the handling of the "offDx" record field, which is described in the following way in the format specification: --- cut --- offDx (4 bytes): A 32-bit unsigned integer that specifies the offset to an intercharacter spacing array, in bytes, from the start of the record in which this object is contained. This value MUST be 32-bit aligned. --- cut --- The offset is supposed to address an array of "Chars" (another field in the text records, specifying the number of characters to be displayed) double words, taking up a total of 4 * N bytes. However, instead of verifying that the provided record is sufficiently large to contain 4 * N bytes at the specified offset, it only checks if it can fit 4 bytes (completely ignoring the actual number of characters in the message, which can be larger than 1). A pseudo-code of the current, vulnerable code is shown below: --- cut --- if ( record_size - offString >= nChars && (!nChars || record_size - 4 >= record->emrtext.offDx) ) { // Validation passed, continue processing the record. } --- cut --- There is definitely a flaw in the implementation, but one which would typically only lead to an out-of-bound read condition, since it's a problem with the sanitization of an input buffer. However, the logic found in the remainder of the function is as follows: - Attempt to convert the textual ANSI string in the record to a wide-char string, using the MultiByteToWideChar() function and the code page specified in the most recently selected font. - If the number of characters converted is equal to the number of bytes in the input buffer, CEmfPlusEnumState::PlayExtTextOut() is called and the function returns. - Otherwise, the function proceeds to rewrite the offDx buffer by calling EmfEnumState::CreateCopyOfCurrentRecord() to allocate an exact copy of the current record (with the same size), and then copying entries of the intercharacter spacing array, omitting those corresponding to bytes which cause the IsDBCSLeadByteEx() function to return true. Once the rewriting is performed, CEmfPlusEnumState::PlayExtTextOut() is called with the new record as the parameter. In order to trigger the more interesting array rewriting behavior, we must get the MultiByteToWideChar() function to convert fewer characters than there are bytes in the input buffer, which means we have to utilize a string in a non-standard encoding, which supports double-byte character sets (DBCS). Luckily, this is possible by selecting a font with an appropriate charset (e.g. SHIFTJIS_CHARSET) into the HDC, and invoking either of the *TextOutA() handlers with a byte stream containing so-called lead bytes (which folds two bytes into a single character, decreasing the eventual return value of the MultiByteToWideChar() call). Since the spacing array in the new record is too small to store entries for all "Chars" characters, it is overflown with data read from memory after the original record buffer. Considering the complexity of the EMF format, other records in the picture file could be easily used to massage the heap such that the record copy is overflown with fully controlled data. The issue has been reproduced in Microsoft Office 2013, as well as a simple C++ program which boils down to the following calls: --- cut --- Graphics graphics(hdc); Metafile *mf = new Metafile(L"C:\\path\\to\\poc.emf"); INT conversionSuccess; mf->ConvertToEmfPlus(&graphics, &conversionSuccess, Gdiplus::EmfTypeEmfPlusDual, NULL); --- cut --- An example crash log from PowerPoint 2013, indicating heap corruption, is shown below (the condition can also be reproduced reliably by enabling Page Heap on the GDI+ client process): --- cut --- (2a8c.2bd8): Break instruction exception - code 80000003 (first chance) eax=00000000 ebx=00000000 ecx=772336ab edx=0022cb85 esi=03bd0000 edi=1171ffc0 eip=7728e815 esp=0022cdd8 ebp=0022ce50 iopl=0 nv up ei pl nz na pe nc cs=0023 ss=002b ds=002b es=002b fs=0053 gs=002b efl=00200206 ntdll!RtlReportCriticalFailure+0x29: 7728e815 cc int 3 0:000> kb ChildEBP RetAddr Args to Child 0022ce50 7728f749 c0000374 772c4270 0022ce94 ntdll!RtlReportCriticalFailure+0x29 0022ce60 7728f829 00000002 64dc1326 03bd0000 ntdll!RtlpReportHeapFailure+0x21 0022ce94 7724ab46 0000000c 03bd0000 1171ffc0 ntdll!RtlpLogHeapFailure+0xa1 0022cf84 771f3431 00000258 00000260 03bd00c4 ntdll!RtlpAllocateHeap+0x7b2 0022d008 695071ec 03bd0000 00000000 00000258 ntdll!RtlAllocateHeap+0x23a 0022d01c 6951bbf1 00000258 116b5104 03bdd558 gdiplus!GpMalloc+0x16 0022d030 69557185 116b50e0 116b50e0 03bdd558 gdiplus!GpGraphics::Save+0x11 0022d4b0 69557bdc 116b50e0 116b5104 116b30d8 gdiplus!CEmfPlusEnumState::PlayExtTextOut+0xda 0022d4ec 69557f25 00000053 03bdae00 00006044 gdiplus!CEmfPlusEnumState::ExtTextOutA+0x136 0022d500 695286ca 00000053 00006044 0d67b568 gdiplus!CEmfPlusEnumState::ProcessRecord+0x13b 0022d51c 69528862 00000053 00000000 00006044 gdiplus!GdipPlayMetafileRecordCallback+0x6c 0022d544 768155f4 9d211b17 0d567180 0d67b568 gdiplus!EnumEmfDownLevel+0x6e 0022d5d0 6952aa36 9d211b17 403581b3 695287f4 GDI32!bInternalPlayEMF+0x6a3 0022d608 6952d199 9d211b17 05462305 0122d674 gdiplus!MetafilePlayer::EnumerateEmfRecords+0x104 0022d6b0 6952f455 00000000 05462305 0022d7d8 gdiplus!GpGraphics::EnumEmf+0x391 0022d810 69534742 00000000 42901225 42901d0b gdiplus!GpMetafile::EnumerateForPlayback+0x7b9 0022d90c 695347c6 03bd2fd8 00000000 00000000 gdiplus!GpGraphics::DrawImage+0x3f5 0022d970 6952c792 03bd2fd8 0022d9e4 0022d9e4 gdiplus!GpGraphics::DrawImage+0x51 0022d9a8 6952ea7a 03bd2fd8 0022d9e4 00000004 gdiplus!GpGraphics::DrawMetafileSplit+0x1f 0022d9fc 6952f4d5 03bdc438 0022dadc 00000000 gdiplus!GpMetafile::ConvertToEmfPlus+0x1c1 0022da20 69504f71 03bdc438 0022dadc 00000004 gdiplus!GpMetafile::ConvertToEmfPlus+0x1d 0022da5c 54793044 03bdc438 03bd2fd8 0022dadc gdiplus!GdipConvertToEmfPlus+0xbf WARNING: Stack unwind information not available. Following frames may be wrong. 0022daf0 548c7b8d 00000000 03bdc438 b93aea31 oart!Ordinal3385+0x7e8 0022df18 548c749b 0022e3a4 094c4380 0022e18c oart!Ordinal655+0x874 0022e12c 54793cbb 0022e3a4 094c4380 0022e18c oart!Ordinal655+0x182 0022e1c0 546bf722 0022e3a4 094c4380 00000000 oart!Ordinal5891+0xad1 0022e200 5474987d 0022e3a4 0d4f7f34 0022e2ec oart!Ordinal3910+0xfd6 0022e214 546bf6b4 0022e3a4 b93ad771 0d4f7f34 oart!Ordinal10880+0x98 0022e258 546beea2 1c0e82b0 b93ad1a5 0d2bce4c oart!Ordinal3910+0xf68 0022e48c 546be7e4 0022e968 0022ed6c 00000002 oart!Ordinal3910+0x756 0022e550 546be4d3 0d2bce48 0022e964 09661440 oart!Ordinal3910+0x98 0022e574 546be440 0022e968 00000002 0022e9b8 oart!Ordinal8924+0xaf 0022e598 546be3aa 0022e968 00000002 0022e9b8 oart!Ordinal8924+0x1c 0022e728 546bc00d 0d83a888 00000000 00000000 oart!Ordinal5363+0x261 0022e784 5474c3c6 00000000 00000000 0d43e458 oart!Ordinal8822+0x20 0022e894 5474c224 0022e964 0022eaa0 00000000 oart!Ordinal5408+0x4f1 0022ea64 5474bff6 0d371f40 0022eaa0 00000000 oart!Ordinal5408+0x34f 0022eb28 54749818 0d371f40 0022ebac 0022eb4c oart!Ordinal5408+0x121 0022eb5c 5473ea78 0d371f40 0022ebac 00000000 oart!Ordinal10880+0x33 0022ed0c 54741fc8 0d371f40 0022ef28 00000000 oart!Ordinal1852+0x241 0022ed44 547425e5 0d371f40 0022ef28 00000000 oart!Ordinal2425+0x5ea 0022ef6c 54743796 0d1a15a0 00000000 0022f34c oart!Ordinal2425+0xc07 0022f0e4 54741d5c 0022f1f0 0473c1ab 3feab68a oart!Ordinal2081+0x292 0022f210 547439d6 0022f2d0 0473c1ab 3feab68a oart!Ordinal2425+0x37e 0022f268 554ecfaa 0022f2d0 0473c1ab 3feab68a oart!Ordinal8518+0xb6 0022f380 554edbd7 b93ac69d 0d3d99bc 0d3d9998 ppcore!PPMain+0x74eff 0022f3b4 554edba9 55497d99 0022f3df b93ac6d9 ppcore!PPMain+0x75b2c 0022f3f0 55497d5a 0022f428 0fabe376 0d3d99b8 ppcore!PPMain+0x75afe 0022f3f8 0fabe376 0d3d99b8 0d184d04 0fabe203 ppcore!PPMain+0x1fcaf 0022f428 0fabd28d 003f9a38 003f7e00 003ff518 mso!Ordinal8295+0x22d 0022f440 0fbd483a 003f9a38 01a81a32 003ff608 mso!Ordinal4996+0x12b 0022f478 0fbd476e 00000001 003ff608 003f7d5c mso!Ordinal3599+0xaf 0022f4d0 0fbce774 003f7d5c 00000000 003f7e9c mso!Ordinal9018+0x334 0022f4ec 0fbcc03c 00000000 0022f55c 00000100 mso!Ordinal8480+0x29d 0022f500 0fbcbf08 003f7e9c 0022f528 5549d3f5 mso!Ordinal4921+0x4c1 0022f50c 5549d3f5 03cd02a0 ffffffff 5549d38b mso!Ordinal4921+0x38d 0022f528 5549d26c 0022f55c 00000001 00000000 ppcore!PPMain+0x2534a 0022f540 5549d238 0022f55c b93ac2b5 01033034 ppcore!PPMain+0x251c1 0022f79c 554780fc 0022f7b8 b93acd25 01033034 ppcore!PPMain+0x2518d 0022f80c 01031572 00312c8c 0022f8ac 0103154a ppcore!PPMain+0x51 0022f818 0103154a 01030000 00000000 00312c8c POWERPNT+0x1572 0022f8ac 76a5338a fffde000 0022f8f8 771f9902 POWERPNT+0x154a 0022f8b8 771f9902 fffde000 64dc254a 00000000 kernel32!BaseThreadInitThunk+0xe 0022f8f8 771f98d5 010312bb fffde000 ffffffff ntdll!__RtlUserThreadStart+0x70 0022f910 00000000 010312bb fffde000 00000000 ntdll!_RtlUserThreadStart+0x1b --- cut --- The poc.emf file is attached. The above analysis was performed using the gdiplus.dll file found in C:\Windows\winsxs\x86_microsoft.windows.gdiplus_6595b64144ccf1df_1.1.7601.23407_none_5c02a2f5a011f9be\GdiPlus.dll on a fully patched Windows 7 64-bit operating system (md5sum c861ee277cd4e2d914740000161956ef). Proof of Concept: https://gitlab.com/exploit-database/exploitdb-bin-sploits/-/raw/main/bin-sploits/40257.zip