AI Security Testing: LLM-02 Finding Sensitive Data Leaks

About this course

Large Language Models are changing how software is built. They are also creating new failure paths that traditional QA workflows were never designed to test.

This course focuses on OWASP LLM-02: Sensitive Information Disclosure, one of the most important risks in the OWASP Top 10 for Large Language Model Applications. The problem is simple to describe and difficult to test correctly: can an LLM-based system reveal data the user should not be allowed to see?

This course is designed for QA engineers, SDETs, automation engineers, and technical testers who want practical testing methods instead of theory alone.

You will learn how sensitive data leaks happen through:

• Runtime context and live session data
• Retrieval and authorization failures
• Aggregate inference and minimum-group threshold failures
• Training-data memorization behavior

You will learn how to:

• Design structured LLM-02 test plans
• Build hypothesis-driven test cases
• Define clear fail signals
• Separate real disclosures from hallucinations
• Capture reproducible evidence
• Write actionable bug reports
• Verify fixes through repeatable testing and regression coverage

The course includes hands-on demonstrations using synthetic data and realistic scenarios. You will see practical QA workflows that can be applied directly to AI-enabled products.

By the end of the course, you will have a repeatable process for testing sensitive information disclosure risks and converting confirmed failures into long-term regression coverage.

This course builds on the same QA-first approach used throughout the OWASP LLM course series.

AI Usage DisclosureAI tools were used during course development to support brainstorming, editing, structure review, and iterative content refinement. All technical content,

Prices, discounts and availability are set by Udemy. We may earn a commission when you purchase through links on this site.