Security of Machine Learning Systems

Explore the security aspects of machine learning systems in this 53-minute OWASP Foundation talk by Shain Singh, Principal Security Architect at F5. Gain clarity on AI, ML, Large Language Models, and Generative AI concepts while focusing on the security implications of these technologies in corporate environments. Discover how application security principles apply to ML systems, including protection from adversarial attacks and operational hygiene through Machine Learning Operations (MLOps). Learn how to adapt white box and black box penetration testing skills to ML systems through practical demonstrations. Witness scenarios showcasing manipulation of image recognition systems, crafting deceptive input sequences for natural language processing models, altering sentiment analysis algorithms, and using steganography to hide malware. Ideal for application security practitioners new to machine learning, this talk bridges the gap between traditional security practices and the evolving landscape of ML system security.