T-MLA: A targeted multiscale log-exponential attack framework for neural image compression

TL;DR

This paper introduces T-MLA, a novel targeted multiscale attack framework in the wavelet domain that significantly degrades neural image compression quality while remaining visually imperceptible, exposing security vulnerabilities.

Contribution

It presents the first targeted multiscale log-exponential attack framework for NIC, improving stealth and effectiveness over existing pixel-space adversarial methods.

Findings

T-MLA causes large quality drops in NIC reconstructions

Perturbations remain visually imperceptible with higher PSNR/VIF

Outperforms PGD-style baselines at similar attack success rates

Abstract

Neural image compression (NIC) has become the state-of-the-art for rate-distortion performance, yet its security vulnerabilities remain significantly less understood than those of classifiers. Existing adversarial attacks on NICs are often naive adaptations of pixel-space methods, overlooking the unique, structured nature of the compression pipeline. In this work, we propose a more advanced class of vulnerabilities by introducing T-MLA, the first targeted multiscale log-exponential attack framework. We introduce adversarial perturbations in the wavelet domain that concentrate on less perceptually salient coefficients, improving the stealth of the attack. Extensive evaluation across multiple state-of-the-art NIC architectures on standard image compression benchmarks reveals a large drop in reconstruction quality while the perturbations remain visually imperceptible. On standard NIC benchmarks, T-MLA achieves targeted degradation of reconstruction quality while improving perturbation imperceptibility (higher PSNR/VIF of the perturbed inputs) compared to PGD-style baselines at comparable attack success, as summarized in our main results. Our findings reveal a critical security flaw at the core of generative and content delivery pipelines.