A hybrid solution for constraint devices to detect microarchitectural attacks - CEA - Commissariat à l’énergie atomique et aux énergies alternatives Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

A hybrid solution for constraint devices to detect microarchitectural attacks

Résumé

We are seeing an increase in cybersecurity attackson resource-constrained systems such as the Internet ofThings (IoT) and Industrial IoT (IIoT) devices. Recently,a new category of attacks has emerged called microarchi-tectural attacks. It targets hardware units of the systemssuch as processor or memory and is often complicated ifnot impossible to remediate since it imposes modifying thehardware. In default to remediation, some solutions proposeto detect these attacks. Yet, most of them are not suitablefor embedded systems since they are based on a complexmachine learning algorithms.In this paper, we propose a security solution for edge-computing attack detection that uses a local-remote machinelearning implementation to find an equilibrium betweenaccuracy and decision-making latency while addressing thememory, performance, and communication bandwidth con-straints of resource-constrained systems. We demonstrateeffectiveness in the detection of multiple microarchitecturalattacks such as Rowhammer or cache attacks on an embed-ded device with an accuracy of 98.75% and a FPR near 0%. To limit the overhead on the communication bus, theproposed solution filters 99% of the samples during normaloperation.
Fichier principal
Vignette du fichier
apres_Soutenance_Paper_Local_Remote.pdf (890.96 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

cea-04473740 , version 1 (22-02-2024)

Identifiants

Citer

Nikolaos Polychronou, Pierre-Henri Thevenon, Maxime Puys, Vincent Beroulle. A hybrid solution for constraint devices to detect microarchitectural attacks. EuroS&P 2023 - Security of Software/Hardware Interfaces (SILM 2023), located with the 8th IEEE European Symposium on Security and Privacy, Jul 2023, Delft, Netherlands. pp.259-269, ⟨10.1109/EuroSPW59978.2023.00033⟩. ⟨cea-04473740⟩
3 Consultations
10 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More