paper-with-me

홈 › Papers

BountyBench: Dollar Impact of AI Agent Attackers and Defenders on Real-World Cybersecurity Systems

2025-05-21 · Andy K. Zhang, Joey Ji, Celeste Menders, Riya Dulepet, Thomas Qin, Ron Y. Wang, Junrong Wu, Kyleen Liao, Jiliang Li, Jinghan Hu, Sara Hong, Nardos Demilew, Shivatmica Murgai, Jason Tran, Nishka Kacheria, Ethan Ho, Denis Liu, Lauren McLane, Olivia Bruvik, Dai-Rong Han, Seungwoo Kim, Akhil Vyas, Cuiyuanxiu Chen, Ryan Li, Weiran Xu, Jonathan Z. Ye, Prerit Choudhary, Siddharth M. Bhatia, Vikram Sivashankar, Yuxuan Bao, Dawn Song, Dan Boneh, Daniel E. Ho, Percy Liang

AI agents have the potential to significantly alter the cybersecurity landscape. To help us understand this change, we introduce the first framework to capture offensive and defensive cyber-capabilities in evolving real-world systems. Instantiating this framework with BountyBench, we set up 25 systems with complex, real-world codebases. To capture the vulnerability lifecycle, we define three task types: Detect (detecting a new vulnerability), Exploit (exploiting a specific vulnerability), and Patch (patching a specific vulnerability). For Detect, we construct a new success indicator, which is general across vulnerability types and provides localized evaluation. We manually set up the environment for each system, including installing packages, setting up server(s), and hydrating database(s). We add 40 bug bounties, which are vulnerabilities with monetary awards from \$10 to \$30,485, and cover 9 of the OWASP Top 10 Risks. To modulate task difficulty, we devise a new strategy based on information to guide detection, interpolating from identifying a zero day to exploiting a specific vulnerability. We evaluate 5 agents: Claude Code, OpenAI Codex CLI, and custom agents with GPT-4.1, Gemini 2.5 Pro Preview, and Claude 3.7 Sonnet Thinking. Given up to three attempts, the top-performing agents are Claude Code (5% on Detect, mapping to \$1,350), Custom Agent with Claude 3.7 Sonnet Thinking (5% on Detect, mapping to \$1,025; 67.5% on Exploit), and OpenAI Codex CLI (5% on Detect, mapping to \$2,400; 90% on Patch, mapping to \$14,422). OpenAI Codex CLI and Claude Code are more capable at defense, achieving higher Patch scores of 90% and 87.5%, compared to Exploit scores of 32.5% and 57.5% respectively; in contrast, the custom agents are relatively balanced between offense and defense, achieving Exploit scores of 40-67.5% and Patch scores of 45-60%.

📄 PDF Abstract BibTeX arXiv:2505.15216

Code (0)

등록된 구현이 없습니다.

Tasks

AI Agent

Methods 이 논문이 사용한 방법론

Attention 설명 없음
Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
Layer Normalization Unlike batch normalization, Layer Normalization directly estimates the normalization statistics from the summed inputs…
Multi-Head Attention 설명 없음
Dense Connections Dense Connections, or Fully Connected Connections, are a type of layer in a deep neural network that use a linear operation where every input is connected to every output…
Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…
Position-Wise Feed-Forward Layer 설명 없음
Absolute Position Encodings Absolute Position Encodings are a type of position embeddings for [Transformer-based models] where positional encodings are…

Similar Papers 제목 키워드 기반

Towards a Multi-Agent Simulation of Cyber-attackers and Cyber-defenders Battles

2025-06-05 · Julien Soulé, Jean-Paul Jamont, Michel Occello, Paul Théron 외

As cyber-attacks show to be more and more complex and coordinated, cyber-defenders strategy through multi-agent approaches could be key to tackle against cyber-attacks as close as entry points in a networked system. This…

IDCAIS: Inter-Defender Collision-Aware Interception Strategy against Multiple Attackers

2021-12-22 · Vishnu S. Chipade, Xinyi Wang, Dimitra Panagou

In the prior literature on multi-agent area defense games, the assignments of the defenders to the attackers are done based on a cost metric associated only with the interception of the attackers. In contrast to that, th…

Multi-Agent Reach-Avoid Games: Two Attackers Versus One Defender and Mixed Integer Programming

2023-09-22 · Hanyang Hu, Minh Bui, Mo Chen

We propose a hybrid approach that combines Hamilton-Jacobi (HJ) reachability and mixed-integer optimization for solving a reach-avoid game with multiple attackers and defenders. The reach-avoid game is an important probl…

Motion Planning

Aerial Swarm Defense using Interception and Herding Strategies

2023-06-04 · Vishnu S. Chipade, Dimitra Panagou

This paper presents a multi-mode solution to the problem of defending a circular protected area (target) from a wide range of attacks by swarms of risk-taking and/or risk-averse attacking agents (attackers). The proposed…

Maintaining Ad-Hoc Communication Network in Area Protection Scenarios with Adversarial Agents

2017-09-04 · Marika Ivanová, Pavel Surynek, Diep Thi Ngoc Nguyen

We address a problem of area protection in graph-based scenarios with multiple mobile agents where connectivity is maintained among agents to ensure they can communicate. The problem consists of two adversarial teams of …