Instructional Leadership, emphasis: K-12 School Leadership (MEd)
Wires that are connected to a computer.
  1. NAU
  2. Research
  3. Innovations
  4. Available technologies
  5. Cybersecurity
  6. Multi-State Physical Unclonable Functions with Machine Learning

Multi-State Physical Unclonable Functions with Machine Learning

Description

This technology uses multi-states to capture the profile of the physical parameters underlining the challenges or responses of physically unclonable functions (PUFs) while only streams of binary bits are downloaded to secure servers. When subjected to natural effects such as aging, temperature changes, bias voltage drifts, or electrostatic interferences, the profile of the PUF challenge-response pairs (CRPs) error rate is predictable when analyzed with multi-states. A machine learning engine (MLE) can compute CRP error rates together with a learning experience that includes the sensing of operating parameters such as temperature or voltage while statistically abnormal challenges are flagged.

Additional information

Patent number and inventor

10,469,273, 10,574,467, and 10,644,892.

Bertrand Cambou and Fatemeh Afghah.

Potential applications

This technology is designed for use with cryptographic systems and authentication methods.

Benefits and advantages

By analyzing the current output of the PUF device in conjunction with the predicted error rates accounting for variations over time, natural drifts in the PUF’s output do not result in errors such as negative authentication, while statistically abnormal challenges may still be caught and flagged as strong negative authentication.

Case number and licensing status

2016-017

This invention is available for licensing.