SEA: Spectral Edge Attack

TL;DR

The paper introduces SEA, a spectral adversarial robustness framework that identifies vulnerable edges in graph-structured data by embedding nodes in a spectrally informed space and scoring edges with the Spade metric. By building a latent-manifold graph and evaluating edge robustness via $L_Y^{+}L_X$, SEA ranks and perturbs the most brittle links to maximize impact with minimal changes. Experiments on graph neural networks and spectral clustering show SEA outperforms random perturbations, underscoring a critical vulnerability in graph-based learning. The work provides a quantitative, edge-centric attack methodology that can guide defense strategies and robustness benchmarking.

Abstract

Graph based machine learning algorithms occupy an important position in today AI landscape. The ability of graph topology to represent complex data structures is both the key strength of graph algorithms and a source of their vulnerability. In other words, attacking or perturbing a graph can severely degrade the performance of graph-based methods. For the attack methods, the greatest challenge is achieving strong attack effectiveness while remaining undetected. To address this problem, this paper proposes a new attack model that employs spectral adversarial robustness evaluation to quantitatively analyze the vulnerability of each edge in a graph under attack. By precisely targeting the weakest links, the proposed approach achieves the maximum attack impact with minimal perturbation. Experimental results demonstrate the effectiveness of the proposed method.