A signature based architecture for Trojan detection

Abstract

Trust in the integrated circuit (IC) fabrication industry is an ongoing concern given the trend towards "fabless" design and associated use of third-parties for fabrication. A Hardware Trojan (HT) introduced during fabrication can corrupt an IC's outputs, leak secret information, and yet go undetected by traditional system testing techniques. In this paper we propose an architecture to detect HTs during IC test or at run-time. An HT that would corrupt an IC's output and otherwise proceed undetected will then be rendered useless by this architecture. This approach will therefore discourage the insertion of HTs in the first place. The proposed architecture takes encryption hardware as a paradigmatic casestudy and uses digital "signatures" derived from the plaintext to identify if the ciphertext has been corrupted by HTs. We test this methodology through simulation on various types of HTs inserted into a lightweight cryptographic system called "PRESENT" [13]. Our results validate that activated HTs are detected by this methodology.

SIMCom: Statistical sniffing of inter-module communications for runtime hardware trojan detection

“…The effects can be catastrophic, such as system failure and leakage of secret encryption keys (e.g., failure of ice-detection module in the P-8A Poseidon [2]), making it imperative to develop effective HT detection techniques. Some of the contemporary HT detection techniques utilize the timing [3] [4], power [5][6] [7], current or electromagnetic signals based golden signatures to detect anomalous electrical behavior [8] [9]. However, in case of the third-party-IP based designs, it is nearly impossible to extract the golden signatures.…”