Death By Python Pickle - Betrayal ML

Explore the critical security vulnerabilities in Python pickle files and their dangerous implications for machine learning systems in this 38-minute conference talk. Learn how pickle files, which allow ML models to "download" knowledge instantly like Neo learning Kung Fu in The Matrix, can be weaponized by attackers to inject malicious code into your AI systems. Discover how threat actors can tamper with these serialized files to create backdoors that execute whenever your model loads, essentially teaching your AI to betray you. Examine real-world examples of pickle-based attacks, understand the technical mechanisms behind these exploits, and explore emerging detection capabilities to protect your ML infrastructure. Gain practical knowledge about supply chain security risks in artificial intelligence and machine learning deployments, with insights from a cybersecurity expert who combines technical expertise with hands-on experience in application security and software supply chain protection.