# Cyber Defense Benchmark: Agentic Threat Hunting Evaluation for LLMs in SecOps

arXiv: https://arxiv.org/abs/2604.19533 · HTML: https://mlsanity.com/paper/2604.19533/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 5 of its 17 references that are in the map.
- First posted on arXiv 21 Apr 2026.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)

## Builds on (most cited first, 5 of 5)

- [InterCode: Standardizing and Benchmarking Interactive Coding with Execution Feedback](https://mlsanity.com/paper/2306.14898/) (2023)
- [Getting pwn'd by AI: Penetration Testing with Large Language Models](https://mlsanity.com/paper/2308.00121/) (2023)
- [SecBench: A Comprehensive Multi-Dimensional Benchmarking Dataset for LLMs in Cybersecurity](https://mlsanity.com/paper/2412.20787/) (2024)
- [ExCyTIn-Bench: Evaluating LLM agents on Cyber Threat Investigation](https://mlsanity.com/paper/2507.14201/) (2025)
- [SIR-Bench: Evaluating Investigation Depth in Security Incident Response Agents](https://mlsanity.com/paper/2604.12040/) (2026)

## Built on by (most cited first, 4 of 4)

- [DiagChain: A Diagnostic Benchmark for Evaluating LLM Agents on Evidence-Grounded Attack Chain Reconstruction](https://mlsanity.com/paper/2608.03591/) (2026)
- [Beyond Success Rate: Cost-Aware Evaluation of Offensive and Defensive Security Agents](https://mlsanity.com/paper/2607.15263/) (2026)
- [Coding Agents Aren't Enough! Evaluating an Enterprise Security Brain for Agentic Cloud Investigations](https://mlsanity.com/paper/2609.30345/) (2026)
- [Operationalizing Cyber Threat Intelligence with GraphRAG](https://mlsanity.com/paper/2608.13050/) (2026)

## Abstract (arXiv)

We introduce the Cyber Defense Benchmark, a benchmark for measuring how well large language model (LLM) agents perform the core SOC analyst task of threat hunting: given a database of raw Windows event logs with no guided questions or hints, identify the exact timestamps of malicious events. The benchmark wraps 106 real attack procedures from the OTRF Security-Datasets corpus - spanning 86 MITRE ATT&CK sub-techniques across 12 tactics - into a Gymnasium reinforcement-learning environment. Each episode presents the agent with an in-memory SQLite database of 75,000-135,000 log records produced by a deterministic campaign simulator that time-shifts and entity-obfuscates the raw recordings. The agent must iteratively submit SQL queries to discover malicious event timestamps and explicitly flag them, scored CTF-style against Sigma-rule-derived ground truth. Evaluating five frontier models - Claude Opus 4.6, GPT-5, Gemini 3.1 Pro, Kimi K2.5, and Gemini 3 Flash - on 26 campaigns covering 105 of 106 procedures, we find that all models fail dramatically: the best model (Claude Opus 4.6) submits correct flags for only 3.8% of malicious events on average, and no run across any model ever finds all flags. We define a passing score as >= 50% recall on every ATT&CK tactic - the minimum bar for unsupervised SOC deployment. No model passes: the leader clears this bar on 5 of 13 tactics and the remaining four on zero. These results suggest that current LLMs are poorly suited for open-ended, evidence-driven threat hunting despite strong performance on curated Q&A security benchmarks.