4 papers · 1 filter
ML Models Are Not Ciphers: Power Analysis for Weight Extraction Is Harder Than You Think
Mohammed Qaiss, Peter Horvath, Michal Shagam +5
Can power-analysis techniques reveal secret weights of machine-learning models? Multiple recent works demonstrate proof-of-concept attacks that can partially recover such weights.…
Physical Side-Channel Attacks on Edge LLMs: Extracting Output Tokens and User Prompts
Najeh Alhalawani, Gunes Acar, Zhuoran Liu +1
The deployment of large language models (LLMs) on edge devices is largely motivated by privacy concerns over sharing sensitive user data with cloud servers. In this paper, we demon…
Triggering Stealthy Feature Map Backdoors via Physical Fault Injection in Embedded Neural Networks
Steyn Hommes, Vincent Dankbaar, Tanguy Stekke +6
Fault injection (FI) attacks on embedded neural network (NN) implementations primarily focus on inducing misclassification by corrupting weights or intermediate computations, overl…
Real-world Edge Neural Network Implementations Leak Private Interactions Through Physical Side Channel
Zhuoran Liu, Senna van Hoek, Péter Horváth +3
Neural networks have become a fundamental component of numerous practical applications, and their implementations, which are often accelerated by hardware, are integrated into all…