COHORT: Collaborative Orchestration for Hardening via Offensive Replay on Emulated Topologies

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Chen Frydman, Aviram Zilberman, Rubin Krief, Abed Showgan, Andres Murillo, Sekiya Motoyoshi, Asaf Shabtai, Yuval Elovici, Rami Puzis
• Venue: arXiv: cs.NI
• Published: 2026-06-29
• Collected: 2026-06-30
• URL: http://arxiv.org/abs/2606.30479v1
• PDF: https://arxiv.org/pdf/2606.30479v1
• DOI: UNKNOWN

Plain-Language Brief

COHORT is an automated end-to-end framework designed to mitigate attacks in enterprise networks, significantly reducing the time and expertise required for this process. It uses a multi-agent workflow, high-fidelity GNS3 emulator, and offensive replay to test candidate mitigations against original attacks, ensuring they disrupt the attack without affecting legitimate network connectivity.

What the Document Says

• COHORT is an automated end-to-end framework for deployable mitigations.
• The framework proposes candidates, implements them as device commands, and refines them through a critique loop on a high-fidelity GNS3 emulator with real vendor firmware (firewall, switch, router).
• Each candidate is evaluated by offensive replay: the original adversary is re-executed on the mitigated network.
• Connectivity-regression check and cumulative evaluation are used for further checks.
• The study involves three topologies and four attack scenarios (ransomware, lateral movement, DNS exfiltration, data theft).
• 46.7% of generated mitigations disrupt the attack and preserve connectivity under replay, which is 4.4 times higher than the single-agent baseline.

Key Concepts Mentioned

• COHORT
• automated mitigation
• multi-agent workflow
• high-fidelity GNS3 emulator
• offensive replay
• connectivity-regression check
• cumulative evaluation

Claims Made by the Document

• COHORT automates the procedure for deployable mitigations.
• 46.7% of generated mitigations both disrupt the attack and preserve connectivity under replay.
• The rate of successful mitigation with COHORT is 4.4 times higher than the single-agent baseline.

Evidence Used

• The text states that COHORT automates the procedure for deployable mitigations and provides data on the success rate.

Methods / Approach

• Role-decomposed multi-agent LLM workflow
• Offensive replay
• Connectivity-regression check
• Cumulative evaluation

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

The framework's success may depend on the quality and variety of attack scenarios tested.

Open Questions

• What is the specific model and tool access used in this study?

Notes for Future Q&A

May include interpretation. Labeled where it does.

• It would be helpful to know the specific details about the model, tools, and baseline for comparison.

Raw Abstract / Excerpt

Mitigating an observed adversary in an enterprise network typically takes weeks of expert work: an analyst derives a mitigation tailored to that adversary, validates it without breaking production, and verifies it disrupts the specific attack. The procedure relies on expert judgment and cannot safely be exercised against the production network. COHORT is the first end-to-end framework to automate this procedure for deployable mitigations. A role-decomposed multi-agent LLM workflow proposes candidates, implements them as real device commands, and refines them through a critique loop, all on a high-fidelity GNS3 emulator running real vendor firmware (firewall, switch, router). Each candidate is evaluated by offensive replay: re-executing the original adversary on the mitigated network for a paired comparison against the unmitigated baseline, rather than the reward-signal or expert-judgment proxies used in prior simulation, hybrid, and configuration-generation work. Two further checks complement replay: a connectivity-regression check (LAN ping and internet HTTP probe) rejects mitigations that disrupt legitimate LAN or internet connectivity, and a cumulative evaluation stacks approved mitigations onto a persistent state to surface compound effects. Across three topologies and four attack scenarios (ransomware, lateral movement, DNS exfiltration, data theft), 46.7% of generated mitigations both disrupt the attack and preserve connectivity under replay, 4.4 times the rate of a single-agent baseline using the same model and tool access. A demo video walking through the framework is available with our released artifacts.