Differences

This shows you the differences between two versions of the page.

Link to this comparison view

isc:labs:05 [2024/11/03 22:10]
florin.stancu
isc:labs:05 [2025/11/04 11:49] (current)
david.gherghita [Exercises]
Line 1: Line 1:
 ====== Lab 05 - Application Security ====== ====== Lab 05 - Application Security ======
 +
 +===== Objectives =====
 +
 +  * Call conventions & stack structure
 +  * Buffer overflow vulnerabilities
 +  * Using pwndbg & pwntools to facilitate exploit development
  
 ===== Resources ===== ===== Resources =====
  
-   ​*[[https://​dhavalkapil.com/​blogs/​Buffer-Overflow-Exploit/​|Buffer overflow explained]] +   * [[https://​dhavalkapil.com/​blogs/​Buffer-Overflow-Exploit/​|Buffer overflow explained]] 
-   ​*[[https://​dhavalkapil.com/​blogs/​Shellcode-Injection/​|Shellcode explained]]+   * [[https://​dhavalkapil.com/​blogs/​Shellcode-Injection/​|Shellcode explained]] 
 +   * [[https://​chatgpt.com/​share/​67279837-b05c-800e-a60a-6629ef3dd7f7|ChatGPT'​s record for stack structure & buffer overflow]] //(same length, but why bother read the opinion of some anonymous industry expert when you got the popular AI kid parroting the same stuff, right?)//
  
 ===== Setup ===== ===== Setup =====
  
-  * [[:​isc:​info:​virtualmachine|Open a lab VM instance]] on [[https://​cloud.grid.pub.ro|OpenStack]], use the **m1.medium** flavor for 2GB of RAM (required by ''​pwndbg''​ :(( ).+  * [[:​isc:​info:​virtualmachine|Open a lab VM instance]] on [[https://​cloud.grid.pub.ro|OpenStack]].
  
-**If you'​re ​not using the OpenStack ​VM**:+**If you're using the local VM for ARM 64 (AArch64)**:
  
-  * Install the 32-bit ​**libc** ​and **gcc-multilib** packages: <​code>​ +  * You have ''​x86''​ cross-compilers pre-installed! Use ''​i686-linux-gnu-gcc''​ to build 32-bit ​programs ​and ''​x86_64-linux-gnu-gcc''​ for AMD64 targets; 
-sudo apt install libc6-dev-i386 gcc-multilib +  * You also have [[https://www.qemu.org/docs/master/user/main.html|qemu-user]] properly installed inside the local VM, so you can simply run any x86 binaries and they'​ll get emulated! 
-</​code>​ +  * If you intend to use ''​objdump''​ aarch64-hosted/emulated x86 binaries, you must use the cross compiler suite'​s,​ so make sure to prefix it, e.g.: ''​x86_86-gnu-linux-objdump''​!
-  * Install the PwnDbg plugin: <​code>​ +
-git clone https://github.com/pwndbg/pwndbg +
-cd pwndbg +
-./setup.sh +
-</code>+
  
-To check if everything is OK, run the command ''​gdb''​ with no arguments. The prompt should be similar to this: +Note that we are using [[https://​github.com/​pwndbg/​pwndbg|PwnDbg]] instead of the classic GDB because it is much more user friendly. Hope you'll like it ;)
- +
-<​code>​ +
-➜ gdb +
-GNU gdb (Ubuntu 12.1-0ubuntu1~22.04) 12.1 +
-... +
-pwndbg>​ +
-</​code>​ +
- +
-Enter ''​q''​ to exit GDB. We are using [[https://​github.com/​pwndbg/​pwndbg|PwnDbg]] instead of the classic GDB because it is much more user friendly. Hope you'll like it ;)+
  
 ===== Overview ===== ===== Overview =====
Line 37: Line 30:
  
 {{:​isc:​labs:​stack_layout.png?​700}} {{:​isc:​labs:​stack_layout.png?​700}}
 +
 +Also check out one of the resources linked on top ^^ !
  
 <note tip> <note tip>
Line 142: Line 137:
     * Hint: you're on 64-bit, check the links above for the calling convention...     * Hint: you're on 64-bit, check the links above for the calling convention...
     * Hint 2: you also don't have debugging info compiled-in,​ so you must use disassembly to find the RBP offset of the ''​buf''​ variable;     * Hint 2: you also don't have debugging info compiled-in,​ so you must use disassembly to find the RBP offset of the ''​buf''​ variable;
 +<spoiler In case of emergency, expand>
 +If this seems too difficult or you wasted too much time, just add ''​-g''​ to the ''​gcc''​ rule inside the Makefile, recompile and try it this way :( 
 +</​spoiler>​
  
 <​solution -hidden> <​solution -hidden>
Line 157: Line 155:
 === [50p] 02. Stack overflow (EZ) === === [50p] 02. Stack overflow (EZ) ===
  
-  * Run & study the ''​buffovf''​ binary. There is a stack overflow ​vulnerability in there, can you see it?+  * Run & study the ''​buffovf''​ binary. There is a vulnerability in there, can you see it?
   * Yep, you **must** use stack overflow to get this flag!   * Yep, you **must** use stack overflow to get this flag!
   * First, try to crash the program. Use programmatically generated input (e.g., from Python3);   * First, try to crash the program. Use programmatically generated input (e.g., from Python3);
Line 170: Line 168:
  
   * Next, try to answer this question: how many bytes do I need to overflow until I get to the EIP saved by the ''​call''​er (also see the stack diagram above)?   * Next, try to answer this question: how many bytes do I need to overflow until I get to the EIP saved by the ''​call''​er (also see the stack diagram above)?
-    * Use either ''​objdump -S''​ or ''​pwndbg>​ disass <​name>''​ on the //​vulnerable function// to figure out the offset of the buffer variable from the EBP register using the assembly code;+    * Use either ''​objdump -S -M intel''​ or ''​pwndbg>​ disass <​name>''​ on the //​vulnerable function// to figure out the offset of the buffer variable from the EBP register using the assembly code;
     * In order to check if the answer is right, try to use an input of the following form: ''​%%b"​A"​ * N + b"​\xEF\xCD\xAB\x98"​%%'';​ this should make the program segfault with the end instruction pointer at ''​0x98ABCDEF''​ (readily visible in pwndbg'​s automatic registers printing);     * In order to check if the answer is right, try to use an input of the following form: ''​%%b"​A"​ * N + b"​\xEF\xCD\xAB\x98"​%%'';​ this should make the program segfault with the end instruction pointer at ''​0x98ABCDEF''​ (readily visible in pwndbg'​s automatic registers printing);
  
   * Things start to become easy; call the ''​for_the_win''​ function (simply replace the address above with the function'​s virtual address)!   * Things start to become easy; call the ''​for_the_win''​ function (simply replace the address above with the function'​s virtual address)!
     * Do not forget: x86 uses little endian encoding for multi-byte integers!     * Do not forget: x86 uses little endian encoding for multi-byte integers!
-    ​Another warning: ASLR is enabled and the program ​is compiled as position independent code! Use ''​gdb''​ to "​leak"​ the function'​s address, which shouldn'​t change while using ''​gdb''​ (it uses the same constant seed for ASLR's randomizer)!+  ​As bonus, can you further chain calls to make the exploited ​program ​gracefully exit?
  
 <​solution -hidden> <​solution -hidden>
 <​code>​ <​code>​
-pwndbg> run "briliantul" < <​(python3 -c '​import sys; sys.stdout.buffer.write(b"​A"​ * 24 + b"​\xa6\x91\x04\x08"​)'​) +pwndbg> run "Salam" < <​(python3 -c '​import sys; sys.stdout.buffer.write(b"​A"​ * 0x19 +b"​\xa6\x91\x04\x08" + b"​\x0e\x93\x04\x08"​ + b"​\xbe\xba\xfe\xca"​)'​)
-... +
-hey, args +
-what's ur last name? +
-almost there, try to supply the correct arguments! +
-bye +
- +
-Program received signal SIGSEGV, Segmentation fault.+
 </​code>​ </​code>​
 </​solution>​ </​solution>​
isc/labs/05.1730664628.txt.gz · Last modified: 2024/11/03 22:10 by florin.stancu
CC Attribution-Share Alike 3.0 Unported
www.chimeric.de Valid CSS Driven by DokuWiki do yourself a favour and use a real browser - get firefox!! Recent changes RSS feed Valid XHTML 1.0