Is Mamba Reliable for Medical Imaging?

TL;DR

This paper assesses the robustness of the Mamba state-space model for medical imaging against various adversarial and corruption attacks, highlighting vulnerabilities and the need for defenses.

Contribution

It provides a comprehensive evaluation of Mamba's robustness under realistic threat models, which was previously underexplored.

Findings

Mamba is vulnerable to input-level adversarial attacks.

Corruption and fault injections significantly reduce accuracy.

Defenses are necessary for reliable deployment.

Abstract

State-space models like Mamba offer linear-time sequence processing and low memory, making them attractive for medical imaging. However, their robustness under realistic software and hardware threat models remains underexplored. This paper evaluates Mamba on multiple MedM-NIST classification benchmarks under input-level attacks, including white-box adversarial perturbations (FGSM/PGD), occlusion-based PatchDrop, and common acquisition corruptions (Gaussian noise and defocus blur) as well as hardware-inspired fault attacks emulated in software via targeted and random bit-flip injections into weights and activations. We profile vulnerabilities and quantify impacts on accuracy indicating that defenses are needed for deployment.