Close

Presentation

CANCELED - SecGraph: Security Oriented Graph NeuRal Network Assisted Protection on Hardware
DescriptionArtificial intelligence (AI) techniques have been developed to automatically ensure intellectual property (IP)-level hardware design's trustworthiness. However, existing works often process the abstracted hardware model, such as netlist level data-flow graph, which leads to inaccurate security analysis and scalable issues. SecGRAPH is proposed in this paper to automate identifying and locating RT-level hardware vulnerabilities by developing a graph neural network-based framework. It extracts syntactical and semantic information directly from the hardware description language program. The experimental results on open-source platforms, such as peripheral IPs in Ariane and Opentitan, show the effectiveness of SecGraph.
Event Type
Work-in-Progress Poster
TimeTuesday, July 11th6:00pm - 7:00pm PDT
LocationLevel 2 Lobby
Topics
AI
Autonomous Systems
Cloud
Design
EDA
Embedded Systems
RISC-V
Security