Traffic Analysis Attacks on Wireless NoC-Based SoCs

Network-on-Chip (NoC) enables on-chip communication between diverse cores in modern System-on-Chip (SoC) designs. Despite the advantages of wireless NoCs (WiNoC) over its electrical counterpart, the use of a shared medium during wireless communication introduces a unique set of vulnerabilities. In WiNoCs, the medium access control (MAC) protocol orchestrates traffic flow across wireless hubs, shaping how traffic patterns emerge based on the protocol and the applications. In this paper, we propose a traffic analysis attack on wireless traffic governed by MAC to infer applications running in an SoC. Specifically, this paper the following major contributions. We outline a threat model involving a malicious wireless hub and a colluding application to leak traffic traces. We utilize deep learning to exploit spatiotemporal traffic features of the MAC protocol to infer applications running in the system, posing significant security and privacy risks. Extensive evaluation demonstrates that our proposed traffic analysis attack can infer applications with high accuracy (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$95\%-99\%$</tex>) across diverse WiNoC configurations and applications. We also propose a lightweight countermeasure to defend against traffic analysis attack that uses stochastic scheduling of wireless packets over virtual wireless hubs. Experimental results show that our proposed countermeasure can defend against traffic analysis attacks with minimal overhead.