####################################################################### Luigi Auriemma Application: EViews http://www.eviews.com Versions: <= 7.0.0.1 (aka 7.2) Platforms: Windows Bugs: A] memory corruption B] heap overflow Exploitation: file Date: 19 Sep 2011 Author: Luigi Auriemma e-mail: aluigi@autistici.org web: aluigi.org ####################################################################### 1) Introduction 2) Bugs 3) The Code 4) Fix ####################################################################### =============== 1) Introduction =============== EViews is a software for econometric and statistical analysis. From vendor's website: "Estimation, forecasting, statistical analysis, graphics, simulation, data management, all in a powerful, graphical object-oriented interface." ####################################################################### ====== 2) Bugs ====== -------------------- A] memory corruption -------------------- The program uses a particular function for allocating memory for the arrays used in the WF1 files. In short if the reallocation fails it's possible to write a memory pointer and a NULL in the expected last two positions of the "supposedly" reallocated array so with possibilities of corrupting memory zones almost arbitrarialy: 00B1A2B0 /$ 56 PUSH ESI ; value + 0x32 00B1A2B1 |. 57 PUSH EDI 00B1A2B2 |. 8B7C24 0C MOV EDI,DWORD PTR SS:[ESP+C] 00B1A2B6 |. 8BF1 MOV ESI,ECX 00B1A2B8 |. 56 PUSH ESI 00B1A2B9 |. 8D04BD 00000000 LEA EAX,DWORD PTR DS:[EDI*4] ; value * 4 (unexploitable) 00B1A2C0 |. 50 PUSH EAX 00B1A2C1 |. B9 38F8E600 MOV ECX,EViews7.00E6F838 00B1A2C6 |. E8 4527F7FF CALL EViews7.00A8CA10 ; allocation ... 00A8BB41 |. 897C39 FC MOV DWORD PTR DS:[ECX+EDI-4],EDI ; write4 00A8BB45 |. C70439 00000000 MOV DWORD PTR DS:[ECX+EDI],0 ; write4 ---------------- B] heap overflow ---------------- Heap overflow/corruption (no additional analysis) during the handling of the PRG files: 00B90BF4 |. 8B1496 MOV EDX,DWORD PTR DS:[ESI+EDX*4] 00B90BF7 |. EB 07 JMP SHORT EViews7.00B90C00 ; EDX is controlled 00B90BF9 |> 397A 04 /CMP DWORD PTR DS:[EDX+4],EDI ; EDI is usually zero 00B90BFC |. 74 0E |JE SHORT EViews7.00B90C0C 00B90BFE |. 8B12 |MOV EDX,DWORD PTR DS:[EDX] 00B90C00 |> 85D2 TEST EDX,EDX 00B90C02 |.^75 F5 \JNZ SHORT EViews7.00B90BF9 00B90C04 |. 33C0 XOR EAX,EAX 00B90C06 |> 5F POP EDI 00B90C07 |> 5E POP ESI 00B90C08 |. 5D POP EBP 00B90C09 |. C2 0400 RETN 4 00B90C0C |> 8B42 08 MOV EAX,DWORD PTR DS:[EDX+8] ; get the new value 00B90C0F \.^EB F5 JMP SHORT EViews7.00B90C06 ... 00B295D8 |. E8 EF750600 CALL EViews7.00B90BCC ; the above function 00B295DD |. 85C0 TEST EAX,EAX 00B295DF |. 74 14 JE SHORT EViews7.00B295F5 00B295E1 |. 8378 4C 00 CMP DWORD PTR DS:[EAX+4C],0 00B295E5 |. 74 0E JE SHORT EViews7.00B295F5 00B295E7 |. 8B48 4C MOV ECX,DWORD PTR DS:[EAX+4C] 00B295EA |. 8B01 MOV EAX,DWORD PTR DS:[ECX] 00B295EC |. 6A 00 PUSH 0 00B295EE |. 56 PUSH ESI 00B295EF |. FF90 9C000000 CALL DWORD PTR DS:[EAX+9C] ; code execution The only "obstacle" to the exploitation could be EDI handled at address 00B90BF9 because sometimes it's zero and other times it's a different value but my proof-of-concept and analysis is absolutely far from being optimized so there are for sure ways to control it better and bypassing the comparison. ####################################################################### =========== 3) The Code =========== http://aluigi.org/poc/eviews_1.zip ####################################################################### ====== 4) Fix ====== No fix. #######################################################################