PDN-Pulse: Ubiquitous and Inherent Anomaly Detection with Cross-Layer Interactions of Power Delivery Network
Abstract
The sophistication of today's computing system gives rise to the increasing threats of malicious and coordinated attacks targeting both software and hardware vulnerabilities. We successfully conduct fundamental research on building a holistic anomaly detection framework by leveraging the inherent cross-layer system infrastructure of power delivery network (PDN). PDN is an essential and ubiquitous component in electronic systems. Our method fundamentally relies on detecting the changes of PDN impedance profile induced by anomalies at different levels. To demonstrate its effectiveness, we conducted proof-of-concept experiments on customized System-on-Chip (SoC) development boards and COTS products, both with successful detection results. We have further developed a system-level security-oriented PDN modeling tool (PowerScout) and a systematic benchmarking method (PCBench) to comprehensively and quantitatively evaluate printed circuit board (PCB)-level attacks. Overall, our project has thoroughly explored the potential of a holistic PDN-based approach for system anomaly detection.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 2021
- Accession Number
- AD1146566
Entities
People
- Xuan Zhang
- Yier Jin
Organizations
- Washington University in St. Louis