Adversarial bug reports induced attacker-desired patches in 90% of trials, while the best tested pre-repair filter caught only 47%, exposing a structural weakness in LLM-based automated program repair.
Wolves in the Repository: A Software Engineering Analysis of the XZ Utils Supply Chain Attack
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
The digital economy runs on Open Source Software (OSS), with an estimated 90\% of modern applications containing open-source components. While this widespread adoption has revolutionized software development, it has also created critical security vulnerabilities, particularly in essential but under-resourced projects. This paper examines a sophisticated attack on the XZ Utils project (CVE-2024-3094), where attackers exploited not just code, but the entire open-source development process to inject a backdoor into a fundamental Linux compression library. Our analysis reveals a new breed of supply chain attack that manipulates software engineering practices themselves -- from community management to CI/CD configurations -- to establish legitimacy and maintain long-term control. Through a comprehensive examination of GitHub events and development artifacts, we reconstruct the attack timeline, analyze the evolution of attacker tactics. Our findings demonstrate how attackers leveraged seemingly beneficial contributions to project infrastructure and maintenance to bypass traditional security measures. This work extends beyond traditional security analysis by examining how software engineering practices themselves can be weaponized, offering insights for protecting the open-source ecosystem.
citation-role summary
citation-polarity summary
fields
cs.SE 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Adversarial Bug Reports as a Security Risk in Language Model-Based Automated Program Repair
Adversarial bug reports induced attacker-desired patches in 90% of trials, while the best tested pre-repair filter caught only 47%, exposing a structural weakness in LLM-based automated program repair.