Guess My Weight: Profiled Side-Channel Recovery of Floating-Point Neural-Network Weights

2026-10-03

AI Summary

This study addresses the challenge of recovering IEEE-754 floating-point neural network weights from embedded devices via side-channel attacks, where the vast candidate space renders extraction intractable. We propose a template attack targeting floating-point multiplication, employing multivariate Gaussian templates with Hamming weight leakage modeling. A novel coarse-to-fine hierarchical search mechanism tailored for the structured 32-bit floating-point space is introduced to achieve bit-level precise recovery of full-precision weights. Experimental evaluations on an Arm Cortex-M4 platform using ChipWhisperer-Lite demonstrate that only 171 power traces are required to attain a 99% recovery success rate, while 263 traces yield 100% exact reconstruction. This work overcomes the longstanding difficulty of extracting high-precision floating-point parameters through side-channel analysis.

Abstract

Neural-network parameters deployed on embedded devices may be exposed through physical side-channel leakage during inference. Existing side-channel attacks on floating-point neural-network parameters have often targeted reduced numerical precision, while recovering the complete IEEE-754 representation remains considerably more challenging because of the large and structured 32-bit candidate space. We present a profiled template attack for bit-exact recovery of an IEEE-754 single-precision neural-network weight from power measurements. The attack targets the floating-point multiplication between a known input and a first-layer weight. During profiling, multivariate Gaussian templates are learned from randomized network configurations using Hamming-weight classes of the multiplication result, while the remaining network parameters act as nuisance variables. To efficiently search the structured 32-bit floating-point candidate space, we use a hierarchical coarse-to-fine-to-exact procedure that progressively increases both the numerical and leakage-model resolution. Experiments on a ChipWhisperer-Lite with an Arm Cortex-M4 demonstrate recovery of the exact float32 representation of the target weight. In the evaluated setting, the attack reaches a bit-exact success rate of 99% with 171 traces and 100% from 263 traces onward. These results demonstrate that profiling can enable practical full-precision extraction of floating-point neural-network parameters from physical leakage.

Problem

• side-channel attack
• floating-point weights
• IEEE-754
• neural network security
• power analysis

Innovation

• Side-channel attack
• Template attack
• Floating-point weights
• IEEE-754 recovery
• Hierarchical search