Does Frontier AI Strengthen Cyber Defense? Evidence from Project Glasswing

Classés selon la proximité de leur titre et de leur résumé avec ceux de cet article. Sélectionnez-en pour les lire ensemble.

Les articles les plus proches

1. Très proche Autre approche du problème Pas pertinent

   Does AI Help Cyber Attackers or Defenders? Evidence from Nonpublic Vulnerabilities and Subsequent Attacks

   Tobias Heldt, Matthew Turk, Christoph R. Landolt et al. · 2026 · Pas encore résumé

   Does AI Help Cyber Attackers or Defenders? teste les modèles sur des vulnérabilités non publiques et des attaques ultérieures, plutôt que d’étudier l’accès anticipé des entreprises.

   Résumé scientifique

   The release decision for frontier AI systems increasingly relies on cyber capability benchmarks, yet public vulnerability benchmarks can expose agents to previously published advisories, exploits, and fixes, making it difficult to distinguish prior exposure from capability on unseen vulnerabilities. We evaluate open-weight and proprietary AI models on exploit generation, vulnerability repair and subsequent attacks in five nonpublic software environments, including vulnerabilities we privately disclosed while they remained unpatched. Researcher-developed and reviewed deterministic graders, not LLM judges, determine task scores. Comparisons with 209 disclosed vulnerabilities and cryptographic …

2. Proche Contexte plus large Pas pertinent

   The Safety Exposure Paradox: How Frontier AI Safety Research May Amplify Cyber Risk in Financial Infrastructure

   Abena Primo, Azubike Okpalaeze, Al Amin et al. · 2026 · Proceedings of the 7th National HBCU Blockchain, Fintech & AI Conference · Pas encore résumé

   The Safety Exposure Paradox analyse comment les divulgations de recherches sur la sûreté des modèles peuvent révéler leurs capacités offensives, sans mesurer leurs effets sur les vulnérabilités logicielles.

   Résumé scientifique

   Frontier artificial intelligence laboratories now publish detailed accounts of how their models behave under adversarial pressure: red-team transcripts, capability thresholds, interpretability findings, and responsible-scaling frameworks. These disclosures are widely treated as safety infrastructure, built to keep dangerous capabilities contained and visible to regulators. This paper argues that the same disclosures can function as a second, less examined kind of infrastructure: a map of where a model’s offensive ceiling sits, available to anyone willing to read it. Using a comparative case study of the published capability-disclosure frameworks of Anthropic, OpenAI, and Google DeepMind, thr…

3. Proche Même méthode, autre problème Pas pertinent

   ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks?

   Zhun Wang, Nico Schiller, Hongwei Li et al. · 2026 · Cryptography and Security · preprint, non relu par les pairs · cité 41 fois · Résumé IA disponible

   ExploitGym évalue des agents d’IA capables de transformer des vulnérabilités logicielles en attaques, tandis que l’étude mesure les effets défensifs de l’accès à une IA de pointe.

   Résumé scientifique

   AI agents are rapidly gaining capabilities that could significantly reshape cybersecurity, making rigorous evaluation urgent. A critical capability is exploitation: turning a vulnerability, which is not yet an attack, into a concrete security impact, such as unauthorized file access or code execution. Exploitation is a particularly challenging task because it requires low-level program reasoning (e.g., about memory layout), runtime adaptation, and sustained progress over long horizons. Meanwhile, it is inherently dual-use, supporting defensive workflows while lowering the barrier for offense. Despite its importance and diagnostic value, exploitation remains under-evaluated. To address this g…

4. Proche Contexte plus large Pas pertinent

   Defensive Sufficiency in a Stackelberg Model of AI Security

   Subhabrata Majumdar, Rajlakshmi Chavan · 2026 · Pas encore résumé

   Defensive Sufficiency formalise les conditions où la découverte et la réparation des failles protègent un système, plutôt que d’estimer les effets d’un accès à l’IA.

   Résumé scientifique

   Feedback from automated testing, human red teaming, and incident response can strengthen an AI system's defenses when discovered failures lead to effective repairs. We study when this feedback process provides sufficient protection and when investing in it is economically worthwhile. We begin by showing that an attack surface composed of finite number of inputs is defended with probability 1 if every unresolved attack has a persistent chance of discovery, repairs are effective, and subsequent updates preserve earlier protection. We derive completion-time bounds and extend the analysis to growing attack surfaces, repairs that generalize across related attacks, and multiple discovery mechanism…

5. Proche* Même question Pas pertinent

   Adopting AI, inviting risk? An empirical analysis of corporate cybersecurity consequences in IT firms

   Wei Xiao, Shan Xu, Jin Li et al. · 2026 · Electronic Markets · Pas encore résumé

   Adopting AI, inviting risk? examine les conséquences de l’adoption de l’IA sur la cybersécurité des entreprises informatiques.

6. Apparenté Même méthode, autre problème Pas pertinent

   CyberDual: AI Attackers versus AI Code Reviewers in Real Software Maintenance

   Chenyuan Yang, Hwiwon Lee, Jongseong Kim et al. · 2026 · Pas encore résumé

   CyberDual évalue des agents d’IA qui détectent et réparent des failles introduites dans du code réel, plutôt que les effets d’un accès anticipé en entreprise.

   Résumé scientifique

   Open-source projects now let AI agents review incoming code, but the code under review may itself come from an AI attacker. We present CyberDual, an arena for this dual role of AI: attacker agents hide an exploitable bug in real maintenance work on widely used open-source projects, and reviewer agents try to find and repair it. Execution decides each match: we rerun the attacker's proof of concept (PoC) on the code that comes out of review. Every attack is new, and the windows can be rebuilt each quarter from the latest upstream history. Three agentic reviewers catch 85-91% of 1,088 valid attacks from six attackers. Yet when the attacker spreads the bug over real upstream commits, about one …

7. Apparenté Article de revue Pas pertinent

   Ten AI Stepping Stones for Cybersecurity

   Ricardo Morla · 2019 · Cryptography and Security · preprint, non relu par les pairs · cité 3 fois · Pas encore résumé

   Ten AI Stepping Stones for Cybersecurity fait le point sur les usages et les risques de l’IA en cybersécurité.

   Résumé scientifique

   With the turmoil in cybersecurity and the mind-blowing advances in AI, it is only natural that cybersecurity practitioners consider further employing learning techniques to help secure their organizations and improve the efficiency of their security operation centers. But with great fears come great opportunities for both the good and the evil, and a myriad of bad deals. This paper discusses ten issues in cybersecurity that hopefully will make it easier for practitioners to ask detailed questions about what they want from an AI system in their cybersecurity operations. We draw on the state of the art to provide factual arguments for a discussion on well-established AI in cybersecurity issues…

8. Apparenté Même méthode, autre problème Pas pertinent

   CWE-bench: Refocusing Agentic Security from Exploitation to Defense

   Gonzalo Gonzalez-Pumariega, Ajay Jagannath, Adit Jain et al. · 2026 · Pas encore résumé

   CWE-bench évalue des agents qui auditent et réparent des failles, alors que l’étude examine l’effet d’un accès à l’IA sur les divulgations et leur exploitation.

   Résumé scientifique

   Cybersecurity evaluations of AI agents have focused largely on offense, yet a working exploit does not show that an agent can make software safe. An attacker needs to exploit only one vulnerability; a defender must find and fix all of them. We introduce CWE-bench, a private benchmark of 120 held-out repository audit-and-repair tasks spanning 73 CWE types, eight languages, and all OWASP Top 10 2025 categories. Agents are given a repository to audit and a description of its threat model, but are not told the weakness, where it occurs, or how many vulnerable sites there are. Each patch must block the exploit while preserving original functionality, as checked by a programmatic verifier. We vali…

Les travaux antérieurs sur lesquels il s’appuie

1. Apparenté Même méthode, autre problème Pas pertinent

   CyberGym: Evaluating AI Agents' Real-World Cybersecurity Capabilities at Scale

   Zhun Wang, Tian-Neng Shi, Jing-Xuan He et al. · 2025 · International Conference on Learning Representations · cité 112 fois · Pas encore résumé

   CyberGym mesure les capacités d’agents à analyser des vulnérabilités réelles, et l’étude cite ce benchmark dans son analyse des effets de l’IA.

   Résumé scientifique

   AI agents have significant potential to reshape cybersecurity, making a thorough assessment of their capabilities critical. However, existing evaluations fall short, because they are based on small-scale benchmarks and only measure static outcomes, failing to capture the full, dynamic range of real-world security challenges. To address these limitations, we introduce CyberGym, a large-scale benchmark featuring 1,507 real-world vulnerabilities across 188 software projects. Adjustable to different vulnerability analysis settings, CyberGym primarily tasks agents with generating a proof-of-concept test that reproduces a vulnerability, given only its text description and the corresponding codebas…

2. Apparenté* Même question Pas pertinent

   Does information security attack frequency increase with vulnerability disclosure? An empirical analysis

   A. Arora, Anand Nandkumar, Rahul Telang · 2006 · Inf. Syst. Frontiers · cité 115 fois · Pas encore résumé

   Does information security attack frequency increase with vulnerability disclosure? étudie le lien entre divulgation de vulnérabilités et fréquence des attaques.

3. Apparenté Contexte plus large Pas pertinent

   The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers

   Kevin Cui, Mert Demirer, Sonia Jaffe et al. · 2026 · Management Sciences · cité 212 fois · Pas encore résumé

   The Effects of Generative AI on High-Skilled Work mesure l’effet d’assistants de codage sur la productivité des développeurs, plutôt que sur la sécurité logicielle.

   Résumé scientifique

   This study evaluates the effect of generative artificial intelligence (AI) on software developer productivity via randomized controlled trials at Microsoft, Accenture, and an anonymous Fortune 100 company. These field experiments, run by the companies as part of their ordinary course of business, provided a random subset of developers with access to an AI-based coding assistant suggesting intelligent code completions. Although each experiment is noisy and results vary across experiments, when data are combined across three experiments and 4,867 developers, our analysis reveals a 26.08% increase (standard error: 10.3%) in completed tasks among developers using the AI tool. Notably, less exper…

4. Apparenté* Même question Utilisé en profondeur par cet article Pas pertinent

   Before we knew it: an empirical study of zero-day attacks in the real world

   Leyla Bilge, Tudor Dumitras · 2012 · Conference on Computer and Communications Security · cité 777 fois · Pas encore résumé

   Before we knew it étudie les attaques zero-day dans le monde réel, un phénomène lié au calendrier d’exploitation des vulnérabilités.

5. Apparenté* Contexte plus large Pas pertinent

   Are Markets for Vulnerabilities Effective?

   S. Ransbotham, Sabyasachi Mitra, Jon Ramsey · 2012 · MIS Q. · cité 134 fois · Pas encore résumé

   Are Markets for Vulnerabilities Effective? porte sur les marchés de vulnérabilités, tandis que l’étude observe les signalements externes liés aux primes de bogues.