Introduction I have been working with C all of my professional and student life. There have been times when I had to look a little deeper to understand what is going on with my buggy program. During my experience, I have learned that there are many tools and techniques that one can use to examine an executable and this post is about that. Furthermore, I will cover a bit about how to employ some reverse engineering practices. Code for the Purpose of Testing As an example, I wrote a pretty basic piece of code with some intentional inclusions. There are two global variables and . msga msgb Two user-defined routines and get executed inside the function. allow deny main One conditional call to an external program is executed using . execvp The idea here is to examine the executable this program creates. Find out where the code I wrote lands in the executable and what compiler adds on top of it. Later I'll showcase some basic reverse engineering that can be done by pretending we haven't seen the code. GitHub #include <stdio.h> #include <unistd.h> char *msga = "Allow"; char *msgb = "Deny"; void allow() { printf("%s\n", msga); } void deny() { printf("%s\n", msgb); } int main(int argc, char **argv) { deny(); int runExternal = 0; if (runExternal) { char* lsargs[] = {"ls", "-l", NULL}; execvp("ls", lsargs); } } While dealing with executable codes we'll encounter a lot of hexadecimal values-- I prefer using Python to do quick arithmetic whenever the need arises. Also, some of the output is going to be too big to paste here in the post so I'll link them in the end. Goal Let's compile the program and get our . a.out gcc examinebin.c As we see in the code the execution of the program basically runs the and halts. deny The goal I am creating for myself is that I'll identify the instruction inside the executable and change it to make sure that is called and then 'ls -a` argument. allow is executed with PS: The tools used to do analysis and their output are listed at the bottom of this post for reference. Call Sequence If we look at the , it is very neatly divided into segments and clearly labeled with symbol names. The code we are interested in is the one we wrote but it's nice to know what everything else is. objdump output The short version is every C program needs a routing that marks the start and end of user written code. main C runtime executes within its framework and takes care of all static and runtime dependencies. main The order of execution can be determined very easily by hooking up the executable with and adding a breakpoint to all symbols defined in section and & . gdb .text _init _fini Let's see what happens. breakpoint : _init, _start, deregister_tm_clones, register_tm_clones, __do_global_dtors_aux, frame_dummy, allow, deny, main, __libc_csu_init, __libc_csu_fini, _fini Below is the call sequence labelled by me based on my understanding of the usual meaning of these symbols: // Initialisation _init (argc=1, argv=0x7fffffffdfd8, envp=0x7fffffffdfe8) _start () __libc_csu_init () _init () frame_dummy () register_tm_clones () // User Code main () deny () // we want to call allow and execvp here instead // Deconstruction and finalisation __do_global_dtors_aux () deregister_tm_clones () deregister_tm_clones () _fini () Identification Now that we know what we don't have to explore we can focus on the task at hand, calling to allow and with . ls -a To do that we will specify our goal properly, basically, we want to: Call allow instead of deny. Change flag value to non-zero. runExternal Change to lsargs` "-l" "-a" in To do that we have to know where these values are in binary and then change them manually without disturbing everything else. Replace deny The hexadecimal code calling from output deny objdump 0000000000001189 <allow>: 00000000000011a3 <deny>: 00000000000011bd <main>: 11e4: e8 ba ff ff ff callq 11a3 <deny> 11e9: c7 45 dc 00 00 00 00 movl $0x0,-0x24(%rbp) From the reference, we know that opcode takes the operand (0xffffffba) which is basically the offset from the next instruction . So, it should point to (0x11a3). callq e8 ba ff ff ff 0x11e9 offset = hex(0xffffffba - 0x100000000) # getting the negative value deny_addr = hex(0x11e9 + int(offset, 16)) print(deny_addr) To call instead, we will have to change (0xffffffba) to something that gives (0x1189) instead. allow allow_addr = hex(0x1189) offset = hex(int(hex(int(allow_addr, 16) - 0x11e9), 16) + 0x100000000) print(offset) # 0xffffffa0 -> a0 ff ff ff to in the binary. So all we need to do is change ba a0 Change runExternal This is quite simple, all we need to do is locate the instruction that is putting the value in the flag. mov 11e9: c7 45 dc 00 00 00 00 movl $0x0,-0x24(%rbp) Then change the value to any non-zero one. ref 00 00 00 00 -> 01 00 00 00 Change "-l" We basically want to change the arguments going into function call. execvp In the assembly, we can see the location where the to has been made and there should be or instruction before that to add the argument into the stack. callq execvp push lea Since these values are hardcoded in binaries all we need to do is get the location of and change it to . -l -a 11f6: 48 8d 05 12 0e 00 00 lea 0xe12(%rip),%rax # 200f <_IO_stdin_used+0xf> 11fd: 48 89 45 e0 mov %rax,-0x20(%rbp) 1201: 48 8d 05 0a 0e 00 00 lea 0xe0a(%rip),%rax # 2012 <_IO_stdin_used+0x12> 1208: 48 89 45 e8 mov %rax,-0x18(%rbp) 121b: 48 8d 3d ed 0d 00 00 lea 0xded(%rip),%rdi # 200f <_IO_stdin_used+0xf> 1222: e8 69 fe ff ff callq 1090 <execvp@plt> The instruction is basically calculating the effective address which in every case here is an offset to the next instruction pointer. lea So we have three addresses, which can be calculated or seen in the output as well. objdump print(hex(0xe12 + 0x11fd)) # 0x200f print(hex(0xe0a + 0x1208)) # 0x2012 print(hex(0xded + 0x1222)) # 0x200f From the output we can clearly see that our strings are really there. hexdump 00002000: 0100 0200 416c 6c6f 7700 4465 6e79 006c ....Allow.Deny.l 00002010: 7300 2d6c 0000 0000 011b 033b 5400 0000 s.-l.......;T... Changing the fourth byte from the right will make . 6c -> 61 l->a Changes Let's summarize and do all the necessary changes to the text output provided by utility. xxd Changes for allow 000011e0: 0000 0000 e8(ba) ffff ffc7 45dc 0000 0000 000011e0: 0000 0000 e8(a0) ffff ffc7 45dc 0000 0000 r Changes fo runExternal flag 000011e0: 0000 0000 e8ba ffff ffc7 45dc (00)00 0000 000011e0: 0000 0000 e8ba ffff ffc7 45dc (01)00 0000 Changes for l -> a 00002010: 7300 2d(6c) 0000 0000 011b 033b 5400 0000 00002010: 7300 2d(61) 0000 0000 011b 033b 5400 0000 Create a new Executable Using xxd Utility xxd -r modified-xxd.txt > a2.out Change Permission chmod +x a2.out Run I can tell you that it actually works but it's better to try yourself. The output is: Before Deny After Allow . .. a.out a2.out Tools Let’s review some tools that tell us about the file from the outside. File Utility that gives the file name, file type, and other format-related information. output Sum Get the checksum and number of blocks in the file. Once we do some reverse engineering this output will tell us that the new executable is not genuine. output ldd Gives the list of shared objects required by the executable. output There are some utilities that give a quick peek about the executable if an in-depth examination is not something you need. strings Displays all printable characters and strings in the file. Works on any file, not just executable. output nm Lists all the symbols present in the executable file address map. output Now comes the in-depth analysis of executable, this includes interpreting the machine code into human-readable form and also figuring out a way to edit the file. objdump Using the -d option you can get the detailed version of each section and segment of your executable along with the interpreted assembly instruction. output or xxd hexdump These are plain read-write tools to deal with binary files and not just executables. The reading part creates a text file showing hexadecimal values at each byte and if possible there is a printable version side by side. Any changes to this output text file can be fed back to the tool, which can then create a binary file. I am using xxd for reading and writing the executable here. output Also published on: https://makeall.dev/notepad/examine-executable/
