Impactful Bit-Flip Search on Full-precision Models

TL;DR

This paper introduces Impactful Bit-Flip Search (IBS), a novel method for efficiently identifying and flipping critical bits in full-precision neural networks to significantly degrade their performance, along with a stealthy modification technique.

Contribution

The paper presents IBS, a new efficient approach for locating critical bits in full-precision models, and introduces Weight-Stealth, a method for covert parameter modifications.

Findings

IBS effectively identifies impactful bits in neural networks.

Weight-Stealth maintains parameter distribution while modifying weights.

The methods increase vulnerability and stealthiness of model attacks.

Abstract

Neural networks have shown remarkable performance in various tasks, yet they remain susceptible to subtle changes in their input or model parameters. One particularly impactful vulnerability arises through the Bit-Flip Attack (BFA), where flipping a small number of critical bits in a model's parameters can severely degrade its performance. A common technique for inducing bit flips in DRAM is the Row-Hammer attack, which exploits frequent uncached memory accesses to alter data. Identifying susceptible bits can be achieved through exhaustive search or progressive layer-by-layer analysis, especially in quantized networks. In this work, we introduce Impactful Bit-Flip Search (IBS), a novel method for efficiently pinpointing and flipping critical bits in full-precision networks. Additionally, we propose a Weight-Stealth technique that strategically modifies the model's parameters in a way that maintains the float values within the original distribution, thereby bypassing simple range checks often used in tamper detection.