PHP PDO MySQL mysqlnd Long Password Buffer Overflow RCE

This repository is a multi-case PHP exploitation research corpus rather than a single exploit. It contains five exploit environments: RCE-CVE-2022-31626, RCE-CVE-2024-2961, SBE-CVE-2019-6977, RCE-N1CTF-php_master, and RCE-SecurinetsCTF-I_hate_php, plus a large Reproduce/ tree for building vulnerable images, validating mitigations, and benchmarking two defensive patch sets (hashtable and refcount guards). Core exploit capability: each main exploit script achieves memory disclosure followed by corruption of Zend/PHP internal structures to redirect execution to system/popen/destructor gadgets, ultimately executing shell commands and exposing output through a file such as 1.php. The payloads are basic hardcoded shell commands (uname/date), so maturity is OPERATIONAL rather than framework-grade weaponized. Per-case structure: - RCE-CVE-2022-31626: Python exploit against a PHP page that accepts MySQL connection parameters and JSON input. The environment includes a rogue_sql_server.py listener on TCP/3306 to emulate a MySQL server and induce the vulnerable state. The exploit performs two leaks (heap and zend), probes candidate libc offsets, then writes a command string and pivots execution to libc system(). - RCE-CVE-2024-2961: Python exploit against a PHP page calling md5_file() on attacker-controlled php://filter/iconv input. It abuses the glibc ISO-2022-CN-EXT conversion path to obtain heap/zend leaks and then corrupts structures to call system(). - RCE-N1CTF-php_master: Includes a full vulnerable PHP web app (DataForm) with session-backed serialized state and compression/inflate operations. The exploit drives the /dataform endpoint through add/insert/append/delete actions, leaks heap data from error snapshots, constructs fake Zend structures, and brute-probes libc deltas until command execution succeeds. - RCE-SecurinetsCTF-I_hate_php: Includes a custom PHP extension (juice.so) implementing xorBMP(). The exploit sends crafted base64 BMPs and JSON layout data to trigger memory corruption, derives heap and PHP text base addresses from returned BMP data, forges a HashTable/frame, and redirects execution to zif_popen. - SBE-CVE-2019-6977: A PHP-only sandbox-bypass / info-leak / code-exec chain. The target script itself contains the exploitation primitives, exposing ?leak and ?pwn modes. The Python wrapper fetches leaked addresses, computes a gadget address, triggers the overwrite, and polls 1.php for command output. Repository support content: - Dockerfiles for each case build pinned PHP source revisions with custom hardening/experimental patches applied. - Reproduce/Patch-hashtable and Reproduce/Patch-refcnt contain mitigation patches, verification harnesses, and benchmark Dockerfiles. These are not exploits; they are defensive evaluation artifacts showing whether the included exploits still succeed after patching. - Reproduce/run_reliability_100.sh and run_performance_tests.sh automate repeated exploit runs and performance measurements. - Misc/static_analysis.ql and patch files document research into allocator/hashtable/refcount hardening. Overall purpose: to provide reproducible exploit demonstrations for several PHP memory-corruption cases and challenge-derived targets, alongside experimental mitigations and benchmarking infrastructure. The code is clearly exploit-oriented, not merely detection, and the included web endpoints, local FastCGI backends, rogue MySQL service, and output files are all fingerprintable operational artifacts.

This repository contains a proof-of-concept exploit for CVE-2022-31626, a buffer overflow vulnerability in PHP's pdo_mysql extension when used with the mysqlnd driver. The exploit is implemented in Python (exploit.py) and targets PHP versions 7.4.x < 7.4.30, 8.0.x < 8.0.20, and 8.1.x < 8.1.7. The exploit works by sending a specially crafted POST request to a PHP endpoint (e.g., mysql_admin.php) with a very long password and specific key/value pairs, triggering a buffer overflow that can leak heap addresses or achieve remote code execution (RCE), depending on configuration. The README.md provides a brief description of the vulnerability. The exploit requires the attacker to be able to supply the MySQL host and password to the PHP application, and the application must expose a vulnerable endpoint. The main attack vector is network-based, targeting a web-accessible PHP script. The code is a proof-of-concept and may require adaptation for different environments.