Harmonizing Intra-coherence and Inter-divergence in Ensemble Attacks for Adversarial Transferability

TL;DR

This paper introduces HEAT, a novel ensemble attack method that improves adversarial transferability by harmonizing shared gradient directions and adaptive weighting, addressing key limitations of prior approaches.

Contribution

The paper proposes HEAT, integrating domain generalization with gradient synthesis and dynamic weight balancing to enhance attack transferability.

Findings

HEAT outperforms existing ensemble attack methods across multiple datasets.

The dual modules effectively balance intra-model coherence and inter-model diversity.

Experimental results demonstrate significant improvements in attack success rates.

Abstract

The development of model ensemble attacks has significantly improved the transferability of adversarial examples, but this progress also poses severe threats to the security of deep neural networks. Existing methods, however, face two critical challenges: insufficient capture of shared gradient directions across models and a lack of adaptive weight allocation mechanisms. To address these issues, we propose a novel method Harmonized Ensemble for Adversarial Transferability (HEAT), which introduces domain generalization into adversarial example generation for the first time. HEAT consists of two key modules: Consensus Gradient Direction Synthesizer, which uses Singular Value Decomposition to synthesize shared gradient directions; and Dual-Harmony Weight Orchestrator which dynamically balances intra-domain coherence, stabilizing gradients within individual models, and inter-domain diversity, enhancing transferability across models. Experimental results demonstrate that HEAT significantly outperforms existing methods across various datasets and settings, offering a new perspective and direction for adversarial attack research.