Secure Architectures for AI-Based Software

In this session, you will gain up-to-date practical knowledge on the evaluation of software architectures which includes AI-based software for security and privacy requirements, and how to improve an insecure architecture using protective measures and best practices such as derived from the OWASP AI Security and Privacy Guide and the current top 10 security issues of machine learning systems (OWASP Machine Learning Security Top 10).
We will consider vulnerabilities of machine learning systems such as: Input Manipulation Attack, Data Poisoning Attack, Model Inversion Attack, Membership Inference Attack, Model Theft, AI Supply Chain Attacks, Transfer Learning Attack, Model Skewing, Output Integrity Attack, and Model Poisoning
… and security and privacy goals and techniques such as: Use Limitation and Purpose Specification, Fairness, Data Minimization and Storage Limitation, Transparency, Data Accuracy, Consent.