Speaker:
D. Richard Kuhn
NIST

Abstract:

Safety and security-critical systems require extensive test and evaluation, but existing high assurance test methods are based on structural coverage criteria that do not apply to many black box AI and machine learning components.   AI/ML systems make decisions based on training data rather than conventionally programmed functions.  Autonomous systems that rely on these components therefore require assurance methods that evaluate input data to ensure that they can function correctly in their environments with inputs they will encounter.  Combinatorial test methods can provide added assurance for these systems and complement conventional verification and test for AI/ML.

This talk reviews some combinatorial methods that can be used to provide assured autonomy, including:

• Background on combinatorial test methods
• Why conventional test methods are not sufficient for many or most autonomous systems
• Where combinatorial methods apply
• Assurance based on input space coverage
• Explainable AI as part of validation

  About: Rick Kuhn is a computer scientist in the Computer Security Division at NIST, and is a Fellow of the Institute of Electrical and Electronics Engineers (IEEE). He co-developed the role based access control (RBAC) model that is the dominant form of access control today. His current research focuses on combinatorial methods for assured autonomy and hardware security/functional verification. He has authored three books and more than 200 conference or journal publications on cybersecurity, software failure, and software verification and testing.

  The weekly security seminar has been held every semester since spring of 1992. We invite personnel at Purdue and visitors from outside to present on topics of particular interest to them in the areas of computer and network security, computer crime investigation, information warfare, information ethics, public policy for computing and security, the computing "underground," and other related topics. More info