Mossad: Defeating Software Plagiarism Detection
summary
A tool called MOSSAD automatically rewrites a compilable submission into semantically equivalent variants that evade popular detectors like Moss by breaking fingerprint hashes; across three C corpora 100% of generated files score below the instructors’ suspicion threshold (25%) and in a mass test 30 variants from five bases yield over 90% under 25% similarity. TAs report zero suspicions in a grading study, and a partial countermeasure—comparing compiled assembly—raises MOSSAD similarities to ≈80%.
AI-written summary · about
chapters (6)
related talks
47mJavaScript Obfuscation: It’s All About the P a c k e r s
47mKeynote: Code-Reuse Attacks and Beyond
44mSlime: Automated Anti-Sandboxing Disarmament System
45m"Advanced Malware Threat: handling obfuscated code using..."
27mIdentifying Malicious OSS Across Ecosystems
56mHomoglyph-Based Attacks: Circumventing LLM Detectors
33mYour Silent Software Saboteur: Open Source Malware
27mPlatPal: Detecting Malicious Documents with Platform Diversity