+- Sinisterly (https://sinister.ly)
+-- Forum: Hacking (https://sinister.ly/Forum-Hacking)
+--- Forum: Tutorials (https://sinister.ly/Forum-Tutorials)
+--- Thread: Tutorial Taking advantage of Intel's messup to obfuscate your code (/Thread-Tutorial-Taking-advantage-of-Intel-s-messup-to-obfuscate-your-code)
Taking advantage of Intel's messup to obfuscate your code - VBQL - 10-16-2017
The method was discussed at Blackhat 2017 (https://www.youtube.com/watch?v=KrksBdWcZgQ) so I can't take credit for it, but I still think it's very interesting and useful. Quite surprisingly, I haven't seen a lot of malware capitalize on this yet.
Basically, this researcher decided to look for undocumented opcodes and quirks in x86 processors. He found a lot of opcodes that weren't documented, and he also discovered a good number of bugs in disassembly software.
In x86, you can prepend 0x66 to jmp (0xe8) or call (0xe9), which supposedly changes the size of the address. AMD processors and disassemblers obey this, but Intel processors ignore it. This discontinuity between what the disassembler says and what Intel does allows you to make the disassembly different from the code that's being executed.
I'm going to use the same example as in the other thread:
So this is the code before the modification. We put a nop before the jump so we have space to work.
Code:
│ 0x00400acb 90 nop
│ ┌─< 0x00400acc e98d000000 jmp 0x400b5eWhen we replace the nop with 0x66, the disassembly changes.
Code:
│ └─< 0x00400acb 66e98d00 jmp 0xb5c
│ 0x00400acf 0000 add byte [rax], alNow, it tells us that we are going to 0x5bc (a completely irrelevant address that doesn't mean shit). But when I run this executable on my intel processor, there is no such error.
Here, the effect is reduced to not being able to follow the code. But what if instead of 0x0000 I had a different instruction? What if it's 3 bytes? Then all of the disassembly until the end of the function is incorrect. Maybe even more if it fails to read ret as its own instruction.
The only drawback of this is that AMD processors will obey the instruction and the code will blow up. For this reason, I suggest you implement a processor checker function that would only run your obfuscated code if the processor is intel.
Anyways, here is how you can set this up on every jump instruction of a particular C file:
Code:
gcc -S <your C file>.c
sed -e 's/\tjmp/\t.byte 0x66\n\tjmp/g' -e 's/\tcall/\t.byte 0x66\n\tcall/g' <your C file>.s -i
gcc <obfuscation options etc..> -static <your .s files>And there you go. Now open up your file in a disassembler and look at your function. Expect most of your calls and jumps to point to nowhere. Definitely different from what the code actually does when you run it.
This guys fuzzing methodology for documenting missing opcodes was just fantastic. Depth-first is really clever. Here is the link his sandsifter tool
https://github.com/xoreaxeaxeax/sandsifter/blob/master/sifter.py
RE: Taking advantage of Intel's messup to obfuscate your code - phyrrus9 - 10-18-2017
This is a really neat trick, I like how clever it is. It really demonstrates to the world the level of asinine Intel processors are. I might actually start writing code that utilizes this in an attempt to block code running on inferior Intel processors.
I still think that ROP gadget chains and ret-libc stuff is far cooler for obfuscation, but this is neat too.
RE: Taking advantage of Intel's messup to obfuscate your code - Blink - 10-18-2017
This is interesting, and shows how dumb Intel is sometimes. (Another example: x86 in general) This can certainly give people trouble when trying to reverse-engineer code.
There are more methods to obfusticate code, but this can be used in conjunction with them for some good results.
RE: Taking advantage of Intel's messup to obfuscate your code - phyrrus9 - 10-18-2017
(10-18-2017, 02:19 AM)Ender Wrote: This is interesting, and shows how dumb Intel is sometimes. (Another example: x86 in general) This can certainly give people trouble when trying to reverse-engineer code.
There are more methods to obfusticate code, but this can be used in conjunction with them for some good results.
Intel has this strong urge to keep everything backwards compatible with the 1980s, when they were doing well as a semiconductor company. Their thought is if they can always go back to the 80's with a simple command sent to the keyboard, they will always do well.
RE: Taking advantage of Intel's messup to obfuscate your code - Blink - 10-18-2017
(10-18-2017, 02:45 AM)phyrrus9 Wrote:(10-18-2017, 02:19 AM)Ender Wrote: This is interesting, and shows how dumb Intel is sometimes. (Another example: x86 in general) This can certainly give people trouble when trying to reverse-engineer code.
There are more methods to obfusticate code, but this can be used in conjunction with them for some good results.
Intel has this strong urge to keep everything backwards compatible with the 1980s, when they were doing well as a semiconductor company. Their thought is if they can always go back to the 80's with a simple command sent to the keyboard, they will always do well.
That's called virtual 8086 mode. If we get desktop-speed ARM CPUs, Intel will either be forced to update, or they'll just die.
RE: Taking advantage of Intel's messup to obfuscate your code - phyrrus9 - 10-18-2017
(10-18-2017, 04:25 AM)Ender Wrote:(10-18-2017, 02:45 AM)phyrrus9 Wrote:(10-18-2017, 02:19 AM)Ender Wrote: This is interesting, and shows how dumb Intel is sometimes. (Another example: x86 in general) This can certainly give people trouble when trying to reverse-engineer code.
There are more methods to obfusticate code, but this can be used in conjunction with them for some good results.
Intel has this strong urge to keep everything backwards compatible with the 1980s, when they were doing well as a semiconductor company. Their thought is if they can always go back to the 80's with a simple command sent to the keyboard, they will always do well.
That's called virtual 8086 mode. If we get desktop-speed ARM CPUs, Intel will either be forced to update, or they'll just die.
We actually had desktop speed RISC CPU's, and we still do. The ones of the past were called PowerPC, and the ones of the present are called SPARC
RE: Taking advantage of Intel's messup to obfuscate your code - Blink - 10-18-2017
(10-18-2017, 04:27 AM)phyrrus9 Wrote:(10-18-2017, 04:25 AM)Ender Wrote:(10-18-2017, 02:45 AM)phyrrus9 Wrote: Intel has this strong urge to keep everything backwards compatible with the 1980s, when they were doing well as a semiconductor company. Their thought is if they can always go back to the 80's with a simple command sent to the keyboard, they will always do well.
That's called virtual 8086 mode. If we get desktop-speed ARM CPUs, Intel will either be forced to update, or they'll just die.
We actually had desktop speed RISC CPU's, and we still do. The ones of the past were called PowerPC, and the ones of the present are called SPARC
PPC stopped being used because it couldn't keep up with the x86 CPUs. SPARC is neat, another one is IBM's POWER7
RE: Taking advantage of Intel's messup to obfuscate your code - phyrrus9 - 10-18-2017
(10-18-2017, 04:28 AM)Ender Wrote:(10-18-2017, 04:27 AM)phyrrus9 Wrote:(10-18-2017, 04:25 AM)Ender Wrote: That's called virtual 8086 mode. If we get desktop-speed ARM CPUs, Intel will either be forced to update, or they'll just die.
We actually had desktop speed RISC CPU's, and we still do. The ones of the past were called PowerPC, and the ones of the present are called SPARC
PPC stopped being used because it couldn't keep up with the x86 CPUs. SPARC is neat, another one is IBM's POWER7
No, that's actually not what happened. It could keep up just fine, IBM refused to allow Apple to use 3GHz+ speeds on the chips, so Apple stopped using them altogether, Microsoft then used PPC until 2014.