What does this research mean for the field?

A hybrid CNN-LSTM framework utilizing hardware performance counters can effectively detect and classify multiple types of cache side-channel attacks with high sensitivity and maintain robustness against adversarial perturbations. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

To develop an adversarially robust, leak-free multi-class framework capable of detecting and classifying cache side-channel attacks in real time using hardware performance counters.

August 28, 2026Journal of Cyber Security Technology

A CNN-LSTM framework with adversarial robustness and dynamic counter selection for real-time multi-class cache side-channel attack detection

View Full Paper

Ask AI

Bookmark

Authors

HMHarshal MoonGujarat Technological University AJAmit JoshiGujarat Technological University

Discussion

Member takes

Implication

Computational study demonstrates robust multi-class cache side-channel attack detection using a hybrid neural network, highlighting improved defense against hardware-level data theft.

Key Points

To develop an adversarially robust, leak-free multi-class framework capable of detecting and classifying cache side-channel attacks in real time using hardware performance counters.
Constructed a hybrid CNN-LSTM architecture comprising 240,837 parameters and 44 hardware performance counter features evaluated under a strict temporal block splitting scheme.
Classified benign executions alongside four major attack types: Flush + Reload, Prime + Probe, Flush + Flush, and Evict + Time.
Assessed defense resilience against adversarial perturbations using Fast Gradient Sign Method (FGSM) and 10-step Projected Gradient Descent (PGD), alongside mutual information-based feature selection.
Achieved 78.04% multi-class test accuracy with zero temporal leakage, 84.84% validation accuracy, and a 30-run mean test accuracy of 77.29% ± 0.97%.
Reached 96.0% accuracy, 100.0% attack detection sensitivity, and a 0.8574% false positive rate when evaluated under realistic production priors.
Retained 62.81% classification accuracy under adversarial perturbation attacks generated via FGSM and 10-step PGD.

Cite This Study

Moon et al. (2026) studied this question.

synapsesocial.com/papers/6a91466cd15324a1df3a9fce https://doi.org/10.1080/23742917.2026.2720501

View Full Paper

Ask AI

Bookmark

Implication

Computational study demonstrates robust multi-class cache side-channel attack detection using a hybrid neural network, highlighting improved defense against hardware-level data theft.

A CNN-LSTM framework with adversarial robustness and dynamic counter selection for real-time multi-class cache side-channel attack detection

Discussion

Member takes

Implication

Key Points

Cite This Study

A CNN-LSTM framework with adversarial robustness and dynamic counter selection for real-time multi-class cache side-channel attack detection

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider