Newest papers

Exposing Vulnerability Monoculture Through Systemic Adversarial Robustness

Leo Elmecker-Plakolm, Łukasz Szpruch, Avishek Joey Bose, Matthew Wicker · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.27353 · cited by 0 works (OpenAlex) · open-access PDF

Russian

Под потенциальной уязвимостью системы, устанавливающей уязвимость, лежит все более однородная техническая рамка общих тренировочных трубопроводов, методов дистилляции и данных. Эти общие алгоритмические выборы могут подвергнуть иным образом отличительную модель экосистемы концентрированным рискам, приводящим к уязвимости монокультуры. Тем не менее, текущая практика в оценке уязвимости противника ограничивается индивидуальными результатами уязвимости и скоростями передачи атак, оставляя открытым неисследованный фундаментальный вопрос: сколько атак достаточно для катастрофического широкомасштабного компромисса? В этой статье мы исследуем эмпирическую природу уязвимостей, приводящих к уязвимости в рамках

Machine translation (Meta M2M100); the English original follows.

Abstract

Beneath the apparent diversity of foundation models lies an increasingly homogeneous technical stack of common training pipelines, distillation methods, and data. These shared algorithmic choices can expose an otherwise distinct model ecosystem to concentrated risks leading to a vulnerability monoculture. Yet current practice in assessing adversarial robustness is restricted to individual robustness scores and attack transfer rates, leaving open the unexplored fundamental question: how many attacks suffice for catastrophic widespread compromise? In this paper, we investigate the empirical nature of shared vulnerabilities within a model pool and introduce Systemic Adversarial Robustness (SAR), a novel measure of shared vulnerability of a model pool through the attack diversity needed to exploit it. Specifically, SAR interpolates between single-attack transferability and independent per-model robustness: small values reveal a vulnerability monoculture, whereas large values indicate that failures are dispersed across models. We theoretically formulate SAR as a partial set-cover problem with provable approximation guarantees---enabling the development of scalable algorithms to empirically measure the systemic adversarial robustness. We empirically test the shared vulnerability of vision and large language models and characterize their shared collective vulnerability curve. Specifically, we demonstrate that model pools distilled from DeepSeek-R1 and Claude Opus 4.6 exhibit lower estimated systemic robustness than their corresponding parent pools, revealing a potential systemic cost of frontier-model distillation. Our framework seeks to establish vulnerability monoculture as a measurable target for a more wholistic adversarial evaluation of model populations.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Adversarial system, Robustness (evolution), Computer science, Exploit, Vulnerability (computing), Vulnerability assessment

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange