Securing AI and Advanced Topics

A sudden spike in undetected fraud. Access attempts with digital images on a facial recognition system. A malware detection system misses new variants of previously identified threats…

As cybersecurity methods advance, cybercriminals pivot—even turning the tables on cybersecurity practitioners by using AI-based tools to outwit our defenses. The third course in Dr. Lanier Watkins’ Certificate in AI for Cybersecurity prepares you to protect critical assets from determined, savvy foes and puts you at the forefront of the shift to adaptive security.

After breaking down a real-world use case of using IBM Watson to detect credit card fraud through advanced sampling techniques and ensemble models, Dr. Watkins takes you into the AI vs. AI domain.

You’ll learn how adversarial examples can compromise neural networks with imperceptible “patch attacks,” how GANs generate synthetic data to evade detection systems, and how reinforcement learning agents can methodically discover ways to bypass security controls while preserving malicious functionality.

By understanding both offensive capabilities and defensive countermeasures, you’ll know how to craft robust AI security systems capable of withstanding even the most advanced learning-based threats.

Through hands-on implementation videos and detailed Jupyter notebooks, you’ll build out defensive techniques including statistical anomaly detection, gradient masking, and adversarial training that neutralize these sophisticated threats.

The course concludes with a critical analysis of H.S. Anderson’s Black Hat-presented malware-evasion framework, examining how reinforcement learning enables malware to systematically evade detection systems through strategic PE file modifications.