Insecure deserialization playbook. Use when Java, PHP, or Python applications deserialize untrusted data via ObjectInputStream, unserialize, pickle, or similar mechanisms that may lead to RCE, file access, or privilege escalation.
Detected risks:
An insecure-deserialization attack playbook for authorized security testing, spanning Java, PHP, Python, .NET, and Ruby. It focuses on the practical gap between finding a deserialization sink and landing a working gadget chain. The guidance begins with traffic fingerprinting: recognising serialized data by its magic bytes and encodings across cookies, POST bodies, headers, and message queues (Java's AC ED and its Base64 form, the PHP object pattern, Python pickle protocol markers, and .NET/Ruby signatures). For Java it walks through ysoserial-based exploitation, choosing among the CommonsCollections, Spring, Groovy, and Hibernate chains, and using the DNS-only URLDNS chain as a damage-free confirmation probe before escalating. Framework-specific sections cover Apache Shiro rememberMe (SHIRO-550 / CVE-2016-4437) and its known default keys, WebLogic over the T3 and IIOP protocols and via XMLDecoder (CVE-2017-10271), and the Java RMI registry, alongside a JDK-version matrix explaining when remote class loading is available or must be bypassed. PHP coverage explains the unserialize magic-method chain and PHAR attacks; Python coverage addresses pickle abuse. A companion reference adds a gadget-chain version-compatibility matrix and further formats such as SnakeYAML, Hessian, Kryo, XStream, and .NET ViewState. It is intended for penetration testers and security researchers working under authorization.
Java (ObjectInputStream), PHP (unserialize/PHAR), and Python (pickle) as primary targets, with notes on .NET ViewState and Ruby YAML plus formats like SnakeYAML, Hessian, Kryo, and XStream.
Use the URLDNS gadget chain, which triggers only a DNS lookup and no code execution; a hit on your collaborator/DNSLog domain confirms the sink before escalating.
Primarily ysoserial for Java gadget-chain payloads (including its RMI exploit module), with references to PHPGGC for PHP and ysoserial.net/Blacklist3r for .NET ViewState.
It includes a matrix showing remote class loading via RMI/LDAP works before 8u121, is partially restricted through 8u190, and is blocked from 8u191 onward, after which an LDAP-returns-gadget approach is needed.
It is an offensive/pentesting playbook for authorized assessments; defenders can reuse its fingerprints and CVE references to prioritize patching and detection.
Quick Setup:
.claude/skills/Repository
yaklang/hack-skills