Sample augmentation-based adversarial training method to counter evasive spectre attacks

Zhongkai Tong

doi:10.1016/j.engappai.2025.113634

Sample augmentation-based adversarial training method to counter evasive spectre attacks

2025-12-29

0

PRE

AI

DOI：10.1016/j.engappai.2025.113634

Original

Original request for help

Save

Abstract

En 中文

• A novel Evasive Spectre attack bypasses AI-based HPC detection systems. • Sample-expansion-based adversarial training detects Evasive Spectre. • Combined hardware events reduce required adversarial samples by up to 94%.

Sample augmentation-based adversarial training method to counter evasive spectre attacks

Abstract

Journal

Organization