Event Horizon: Hardware Security Emulators for Next-Generation Edge AI/ML

Abstract

Hardware security concerns such as side-channel analysis, fault injection attacks, and hardware backdoors are emerging threats for artificial intelligence (AI) and machine learning (ML) applications. These attacks can cause stealing a trained AI/ML model, violating the data privacy, or making hazardous misclassifications in critical cyberinfrastructure that rely on AI/ML applications. This project will develop the first emulation framework that will allow comprehensive and rapid hardware security testing of edge AI/ML devices used in cyber and physical systems. Using data from real attacks and the target hardware details, we will establish a hardware security simulation model for popular AI/ML accelerators that closely mimics the events in real hardware when attacked. We will thenuse the developed model to reveal attacks on existing and future neural network topologies. The project will allow rapid and accurate evaluation of risks on new architectures/applications and the quality of defenses.

Document Details

Document Type
DoD Grant Award
Publication Date
Jan 12, 2023
Source ID
N000142312103

Entities

People

  • Aydin Aysu

Organizations

  • North Carolina State University
  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Computer science
  • Mathematics

Readers

  • Cybersecurity.
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Neural Networks
  • Cyber