Defending the Citadel: Fault Injection Attacks against Dynamic Information Flow Tracking and Related Countermeasures

William Pensec (1, 2) , Francesco Regazzoni (3, 4) , Vianney Lapotre (1, 2) , Gogniat Guy (1, 2)

1 Lab-STICC_ARCAD - Equipe Hardware ARchitectures and CAD tools
2 UBS Lorient - Université de Bretagne Sud - Lorient
3 UvA - University of Amsterdam [Amsterdam] = Universiteit van Amsterdam
4 USI - Università della Svizzera italiana = University of Italian Switzerland

Embedded processors are key components of Internet of Things (IoT) devices of Cyber-Physical Systems (CPSs) that manipulate sensitive data. In order to mitigate software attacks, hardware-assisted Dynamic Information Flow Tracking (DIFT) has been integrated into embedded processors. Due to their proximity to attackers, IoT devices are also exposed to physical attacks such as Fault Injection Attacks (FIAs). In this paper, we protect DIFT from fault injection attacks by extending current DIFT support with fault detection and correction capabilities. To do so, we design, implement and evaluate two countermeasures based on parity bit or Hamming code to protect DIFT-related registers of RISC-V CPUs. Our experimental results, obtained using the D-RI5CY processor as a use case, show a 100% fault detection and fault correction when relying on a Hamming code-based protection and a low area overhead (10.6\%) compared to the original design.

Mots clés: Hardware security RISC-V DIFT Fault Injections Attacks Error-Correcting Code Countermeasures; Hardware security; RISC-V; DIFT; Fault Injections Attacks; Error-Correcting Code; Countermeasures

Domaines: Architectures Matérielles [cs.AR]; Cryptographie et sécurité [cs.CR]

William Pensec, Francesco Regazzoni, Vianney Lapotre, Gogniat Guy. Defending the Citadel: Fault Injection Attacks against Dynamic Information Flow Tracking and Related Countermeasures. 2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), Jul 2024, Knoxville, United States. ⟨hal-04620057⟩