Reproduce “Policy-Value Guided MDP-MCTS Framework for Cyber Kill-Chain Inference” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Chitraksh Singh, Monisha Dhanraj, Ken Huang

Researchers introduce a Policy–Value Guided MDP–MCTS framework that infers complete, seven-phase MITRE ATT&CK kill chains directly from unstructured cyber threat intelligence reports. This framework leverages semantic interpretation, symbolic reasoning, and a guided search mechanism to achieve superior behavioral alignment with known threat actor tradecraft compared to a Transformer baseline.

Key insights

  • Combining semantic interpretation of CTI with an MDP for structural dependencies and MCTS for guided search enables the automated reconstruction of behaviorally plausible and coherent kill chains.
  • The use of FiLM modulation allows the global CTI context to dynamically transform candidate technique embeddings, adapting the search space to the specifics of the threat narrative.
  • A multi-objective reward function, incorporating relevance, cohesion, transition probability, stealth, and mitigation penalties, effectively supervises the PVN and guides the MCTS search toward optimal attack paths.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.