ExCyTIn-Bench: Evaluating LLM agents on Cyber Threat Investigation

• URL: http://arxiv.org/abs/2507.14201v2
• Date: Mon, 01 Sep 2025 20:02:00 GMT
• Title: ExCyTIn-Bench: Evaluating LLM agents on Cyber Threat Investigation
• Abstract summary: We present ExCyTIn-Bench, the first benchmark to evaluate an LLM agent x on the task of Cyber Threat Investigation.<n>We leverage security logs extracted with expert-crafted detection logic to build threat investigation graphs.<n>We generate questions with LLMs using paired nodes on the graph, taking the start node as background context and the end node as answer.
• Score: 15.227180564453016
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We present ExCyTIn-Bench, the first benchmark to Evaluate an LLM agent x on the task of Cyber Threat Investigation through security questions derived from investigation graphs. Real-world security analysts must sift through a large number of heterogeneous alert signals and security logs, follow multi-hop chains of evidence, and compile an incident report. With the developments of LLMs, building LLM-based agents for automatic thread investigation is a promising direction. To assist the development and evaluation of LLM agents, we construct a dataset from a controlled Azure tenant that covers 8 simulated real-world multi-step attacks, 57 log tables from Microsoft Sentinel and related services, and 589 automatically generated questions. We leverage security logs extracted with expert-crafted detection logic to build threat investigation graphs, and then generate questions with LLMs using paired nodes on the graph, taking the start node as background context and the end node as answer. Anchoring each question to these explicit nodes and edges not only provides automatic, explainable ground truth answers but also makes the pipeline reusable and readily extensible to new logs. This also enables the automatic generation of procedural tasks with verifiable rewards, which can be naturally extended to training agents via reinforcement learning. Our comprehensive experiments with different models confirm the difficulty of the task: with the base setting, the average reward across all evaluated models is 0.249, and the best achieved is 0.368, leaving substantial headroom for future research. Code and data are coming soon!

Related papers

• SherAgent: Scaling Attack Investigation in the Wild via LLM-Empowered Iterative Query-Filter Backtracking [32.566471484012276]
  Provenance-based attack investigation enables viable automation by standardizing data and query logic.<n>We collaborate with the SOC of a major Internet corporation serving billions of users to develop SherAgent.<n>SherAgent operates with remarkable efficiency, incurring under $0.10 in API costs and requiring less than 4 minutes per investigation.
  arXiv  Detail & Related papers  (2026-07-10T08:11:10Z)
• From Detection to Response: A Deep Learning and Retrieval-Augmented Generation Framework for Network Intrusion Mitigation [0.7136933021609079]
  This paper presents a unified, end-to-end framework that closes the gap between threat detection and actionable response.<n>The system operates in two tightly coupled stages. First, an ensemble of three independently trained binary Deep Neural Networks (DNNs) classifies network traffic flows as Benign, Denial of Service (DoS) or Distributed Denial of Service (DDoS)<n>Second, a Retrieval-Augmented Generation (RAG) pipeline constructs explanation-aware prompts from the top-5 anomalous features.
  arXiv  Detail & Related papers  (2026-05-18T07:17:55Z)
• Heimdallr: Characterizing and Detecting LLM-Induced Security Risks in GitHub CI Workflows [24.281621604693726]
  GitHub Continuous Integration (CI) increasingly integrate Large Language Models (LLMs) to automate review, triage, content generation, and repository maintenance.<n> externally controllable workflow inputs can shape LLM prompts and outputs, which may in turn affect security decisions, repository state, or privileged execution.<n>We present the first study of LLM-induced security risks in GitHub CI.
  arXiv  Detail & Related papers  (2026-05-07T10:16:26Z)
• Towards Agentic Investigation of Security Alerts [0.07646713951724012]
  Security analysts are overwhelmed by the volume of alerts and the low context provided by detection systems.<n>Early-stage investigations typically require manual correlation across multiple log sources.<n>We present an experimental, agentic workflow that leverages large language models (LLMs) augmented with predefined queries.
  arXiv  Detail & Related papers  (2026-04-28T16:52:12Z)
• Towards Compositional Generalization in LLMs for Smart Contract Security: A Case Study on Reentrancy Vulnerabilities [35.39583123277091]
  This paper proposes a post-training algorithm based on atomic task decomposition and fusion.<n>We decompose the reentrancy vulnerability detection task into four linearly independent atomic tasks.<n>By training on synthetic datasets, we generate three compiler-verified datasets.<n>We then employ the Slither tool to extract structural information from the control flow graph and data flow graph.
  arXiv  Detail & Related papers  (2026-01-11T13:52:07Z)
• CORTEX: Collaborative LLM Agents for High-Stakes Alert Triage [10.088447487211893]
  Security Operations Centers (SOCs) are overwhelmed by tens of thousands of daily alerts.<n>This overload creates alert fatigue, leading to overlooked threats and analyst burnout.<n>We propose CORTEX, a multi-agent LLM architecture for high-stakes alert triage.
  arXiv  Detail & Related papers  (2025-09-30T22:09:31Z)
• Guided Reasoning in LLM-Driven Penetration Testing Using Structured Attack Trees [1.2397617816774036]
  Existing Large Language Models (LLMs) for cybersecurity penetration testing rely on self-guided reasoning.<n>We propose a guided reasoning pipeline for penetration testing LLM agents that incorporates a deterministic task tree built from the MITRE ATT&CK Matrix.<n>Our pipeline guided the LLM agent through 71.8%, 72.8%, and 78.6% of subtasks using Llama-3-8B, Gemini-1.5, and GPT-4, respectively.
  arXiv  Detail & Related papers  (2025-09-09T17:19:33Z)
• From Alerts to Intelligence: A Novel LLM-Aided Framework for Host-based Intrusion Detection [16.59938864299474]
  Large Language Models (LLMs) have great potentials to advance the state of host-based intrusion detection system (HIDS)<n>LLMs have extensive knowledge of attack techniques and their ability to detect anomalies through semantic analysis.<n>In this work, we explore the direction of building a customized LLM pipeline for HIDS and develop a system named SHIELD.
  arXiv  Detail & Related papers  (2025-07-15T00:24:53Z)
• LLM Embedding-based Attribution (LEA): Quantifying Source Contributions to Generative Model's Response for Vulnerability Analysis [1.8780665584728753]
  Large Language Models (LLMs) have been widely adopted for cybersecurity threat analysis.<n>Retrieval-Augmented Generation (RAG) pipelines mitigate the problem by injecting up-to-date authoritative sources into the model context.<n>This work proposes LLM Embedding-based Attribution (LEA) -- a novel, explainable metric to paint a clear picture on the 'percentage of influence' the pre-trained knowledge vs. retrieved content has for each generated response.
  arXiv  Detail & Related papers  (2025-06-12T21:20:10Z)
• IDA-Bench: Evaluating LLMs on Interactive Guided Data Analysis [60.32962597618861]
  IDA-Bench is a novel benchmark evaluating large language models in multi-round interactive scenarios.<n>Agent performance is judged by comparing its final numerical output to the human-derived baseline.<n>Even state-of-the-art coding agents (like Claude-3.7-thinking) succeed on 50% of the tasks, highlighting limitations not evident in single-turn tests.
  arXiv  Detail & Related papers  (2025-05-23T09:37:52Z)
• AGENTIF: Benchmarking Instruction Following of Large Language Models in Agentic Scenarios [51.46347732659174]
  Large Language Models (LLMs) have demonstrated advanced capabilities in real-world agentic applications.<n>AgentIF is the first benchmark for systematically evaluating LLM instruction following ability in agentic scenarios.
  arXiv  Detail & Related papers  (2025-05-22T17:31:10Z)
• CyberLLMInstruct: A New Dataset for Analysing Safety of Fine-Tuned LLMs Using Cyber Security Data [2.2530496464901106]
  integration of large language models into cyber security applications presents significant opportunities.<n>CyberLLMInstruct is a dataset of 54,928 instruction-response pairs spanning cyber security tasks.<n>Fine-tuning models can achieve up to 92.50 percent accuracy on the CyberMetric benchmark.
  arXiv  Detail & Related papers  (2025-03-12T12:29:27Z)
• AttackSeqBench: Benchmarking Large Language Models' Understanding of Sequential Patterns in Cyber Attacks [13.082370325093242]
  We introduce AttackSeqBench, a benchmark to evaluate Large Language Models' (LLMs) capability to understand and reason attack sequences in Cyber Threat Intelligence (CTI) reports.<n>Our benchmark encompasses three distinct Question Answering (QA) tasks, each task focuses on the varying granularity in adversarial behavior.<n>We conduct extensive experiments and analysis with both fast-thinking and slow-thinking LLMs, while highlighting their strengths and limitations in analyzing the sequential patterns in cyber attacks.
  arXiv  Detail & Related papers  (2025-03-05T04:25:21Z)
• Scaling Autonomous Agents via Automatic Reward Modeling And Planning [52.39395405893965]
  Large language models (LLMs) have demonstrated remarkable capabilities across a range of tasks.<n>However, they still struggle with problems requiring multi-step decision-making and environmental feedback.<n>We propose a framework that can automatically learn a reward model from the environment without human annotations.
  arXiv  Detail & Related papers  (2025-02-17T18:49:25Z)
• SnipGen: A Mining Repository Framework for Evaluating LLMs for Code [51.07471575337676]
  Language Models (LLMs) are trained on extensive datasets that include code repositories.<n> evaluating their effectiveness poses significant challenges due to the potential overlap between the datasets used for training and those employed for evaluation.<n>We introduce SnipGen, a comprehensive repository mining framework designed to leverage prompt engineering across various downstream tasks for code generation.
  arXiv  Detail & Related papers  (2025-02-10T21:28:15Z)
• DCA-Bench: A Benchmark for Dataset Curation Agents [9.60250892491588]
  Data quality issues, such as incomplete documentation, inaccurate labels, ethical concerns, and outdated information, remain common in widely used datasets.<n>With the surging ability of large language models (LLM), it's promising to streamline the discovery of hidden dataset issues with LLM agents.<n>In this work, we establish a benchmark to measure LLM agent's ability to tackle this challenge.
  arXiv  Detail & Related papers  (2024-06-11T14:02:23Z)
• Get my drift? Catching LLM Task Drift with Activation Deltas [55.75645403965326]
  Task drift allows attackers to exfiltrate data or influence the LLM's output for other users.<n>We show that a simple linear classifier can detect drift with near-perfect ROC AUC on an out-of-distribution test set.<n>We observe that this approach generalizes surprisingly well to unseen task domains, such as prompt injections, jailbreaks, and malicious instructions.
  arXiv  Detail & Related papers  (2024-06-02T16:53:21Z)
• InfiAgent-DABench: Evaluating Agents on Data Analysis Tasks [84.7788065721689]
  In this paper, we introduce InfiAgent-DABench, the first benchmark specifically designed to evaluate LLM-based agents on data analysis tasks. This benchmark contains DAEval, a dataset consisting of 257 data analysis questions derived from 52 CSV files. Building on top of our agent framework, we develop a specialized agent, DAAgent, which surpasses GPT-3.5 by 3.9% on DABench.
  arXiv  Detail & Related papers  (2024-01-10T19:04:00Z)
• MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation [96.71370747681078]
  We introduce MLAgentBench, a suite of 13 tasks ranging from improving model performance on CIFAR-10 to recent research problems like BabyLM. For each task, an agent can perform actions like reading/writing files, executing code, and inspecting outputs. We benchmark agents based on Claude v1.0, Claude v2.1, Claude v3 Opus, GPT-4, GPT-4-turbo, Gemini-Pro, and Mixtral and find that a Claude v3 Opus agent is the best in terms of success rate.
  arXiv  Detail & Related papers  (2023-10-05T04:06:12Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.