DEF CON 33 - Preventing One of The Largest Supply-Chain Attacks in History - Maksim Shudrak

Supply chain attacks via abandoned AWS S3 buckets are a serious, scalable threat enabled by LLMs and big data analysis, with real-world impact on thousands of organizations and government networks.

Key Takeaways:
• Abandoned S3 buckets can be reclaimed and exploited for code execution, affecting thousands of hosts across 158 countries 1:47-2:30.
• Attackers can poison PyTorch models or Maven/PI packages to inject backdoors, leading to code execution on systems from Fortune 500 companies to government agencies 15:00-16:00.
• Malware like RK and Linker, active years ago, can be weaponized via S3 buckets to inject malicious JavaScript into browsers, enabling data exfiltration from sites like Google and Facebook 25:00-27:20.
• The attack surface is vast—4.5K GitHub buckets, 480 Maven/PI buckets, and 188 malware-related buckets were identified, with 28,000 users at risk 10:00-11:00.

This threat is not theoretical; it highlights a critical gap in supply chain security, where dangling bucket references and lack of bucket name predictability enable large-scale attacks. Cloud providers must adopt secure-by-default policies, including non-recyclable bucket names and strict access controls to prevent exploitation.

Sources:

• 1:47 Discussion of supply chain attacks and S3 bucket exploitation.
• 15:00 Demonstration of PyTorch model poisoning and code execution.
• 25:00 Malware weaponization via S3 buckets and JavaScript injection.
• 10:00 Scale of buckets and user impact a