Can We Protect Our Private Data in the Machine Learning Age?

Explore the critical challenges and cutting-edge innovations in protecting private data during the machine learning era in this comprehensive keynote delivered at the Intelligent Systems Conference (IntelliSys) 2025. Learn from Marten van Dijk, IEEE Fellow and founder of the Computer Security group at CWI, Netherlands, who brings over 20 years of expertise spanning academia and industry, including prestigious roles at Philips Research, RSA Laboratories, MIT, and Vrije Universiteit Amsterdam. Discover the evolution of AI and machine learning from early neural networks to modern foundation models, while examining sophisticated techniques in confidential computing including Secure Multi-Party Computation (MPC), partial and fully homomorphic encryption, secure processor architectures, and trusted execution environments (TEEs). Understand the complex challenges of differential privacy and how to balance privacy guarantees with practical utility in machine learning models. Gain insights into emerging paradigms such as PAC Privacy and instance-based cryptography that could revolutionize the field. Examine practical considerations for federated learning and low-resource devices, while navigating the legal and governance landscape including GDPR, AI Act, and risk-based cybersecurity frameworks. Analyze real-world examples of privacy-preserving techniques and their deployment limitations through a unique timeline that connects decades of research in cryptography, secure computation, and privacy with modern AI applications. Perfect for researchers, practitioners, and enthusiasts working in machine learning, AI security, data privacy, federated learning, cryptography, and confidential computing who want to understand current state-of-the-art methods for securing sensitive data and the trade-offs between privacy and model utility.