This episode explores MalGEN, a multi-agent AI framework developed by researchers at IIT Kanpur that autonomously generates novel, functional malware capable of evading modern detection systems. The discussion examines why LLM-generated malware represents a qualitative shift beyond traditional polymorphic and metamorphic techniques — rather than mutating a fixed payload syntactically, LLMs reason about semantic intent and produce entirely new code that achieves the same effect through different computational paths. A central focus is MalGEN's alignment with MITRE ATT&CK tactics, techniques, and procedures, meaning the generated malware maps to documented real-world intrusion patterns rather than merely bypassing signature databases. The hosts pressure-test the paper's red teaming justification — framing MalGEN as a defensive stress-testing tool — while examining its most unsettling capability: automating sandbox-aware, environment-detecting evasion previously requiring nation-state-level expertise. The conversation anchors on the dual-use tension at the core of publishing a reproducible malware generation framework under academic cover.

Sources:

1. MalGEN: A Generative Agent Framework for Modeling Malicious Software in Cybersecurity — Bikash Saha, Sandeep Kumar Shukla, 2025

http://arxiv.org/abs/2506.07586

2. https://arxiv.org/pdf/2510.23883

3. https://arxiv.org/pdf/2601.05293

4. https://arxiv.org/pdf/2508.05674

5. Evaluating the Cybersecurity Capabilities of LLMs: A Comprehensive Study — Bhatt, M., Chennabasappa, S., Nikolaidis, C., et al. (Meta), 2023

https://scholar.google.com/scholar?q=Evaluating+the+Cybersecurity+Capabilities+of+LLMs:+A+Comprehensive+Study

6. From Chatbots to Phishbots? Phishing Scam Generation in Commercial Large Language Models — Heiding, F., Schneier, B., Vishwanath, A., Bernstein, J., Park, P.S., 2024

https://scholar.google.com/scholar?q=From+Chatbots+to+Phishbots?+Phishing+Scam+Generation+in+Commercial+Large+Language+Models

7. PentestGPT: An LLM-Empowered Automatic Penetration Testing Framework — Deng, G., Liu, Y., Mayoral-Vilches, V., et al., 2024

https://scholar.google.com/scholar?q=PentestGPT:+An+LLM-Empowered+Automatic+Penetration+Testing+Framework

8. LLM4Decompile: Decompiling Binary Code with Large Language Models — Tan, Z., Ma, H., Xu, H., et al., 2024

https://scholar.google.com/scholar?q=LLM4Decompile:+Decompiling+Binary+Code+with+Large+Language+Models

9. LLM Agents can Autonomously Exploit One-Day Vulnerabilities — Fang, R., Bindu, R., Gupta, A., Kang, D., Boneh, D., 2024

https://scholar.google.com/scholar?q=LLM+Agents+can+Autonomously+Exploit+One-Day+Vulnerabilities

10. Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned — Ganguli, D., Lovitt, L., Kernion, J., et al. (Anthropic), 2022

https://scholar.google.com/scholar?q=Red+Teaming+Language+Models+to+Reduce+Harms:+Methods,+Scaling+Behaviors,+and+Lessons+Learned

11. Trojan Detection Benchmark (TrojAI): Evaluating Backdoor Defenses on Neural Networks — Karra, K., Ashcraft, C., Fendley, N. (IARPA / Johns Hopkins APL), 2020

https://scholar.google.com/scholar?q=Trojan+Detection+Benchmark+(TrojAI):+Evaluating+Backdoor+Defenses+on+Neural+Networks

12. Evading Machine Learning Malware Detection — Grosse, K., Papernot, N., Manoharan, P., Backes, M., McDaniel, P., 2017

https://scholar.google.com/scholar?q=Evading+Machine+Learning+Malware+Detection

13. Malware Detection by Eating a Whole EXE — Raff, E., Barker, J., Sylvester, J., Brandon, R., Catanzaro, B., Nicholas, C., 2018

https://scholar.google.com/scholar?q=Malware+Detection+by+Eating+a+Whole+EXE

14. Adversarial Malware Binaries: Evading Deep Learning for Malware Detection in Executables — Kolosnjaji, B., Demontis, A., Biggio, B., Maiorca, D., Giacinto, G., Eckert, C., Roli, F., 2018

https://scholar.google.com/scholar?q=Adversarial+Malware+Binaries:+Evading+Deep+Learning+for+Malware+Detection+in+Executables

15. DQEAF: Malware Adversarial Examples Generation with Reinforcement Learning — Fang, Y., Liu, Y., Huang, C., Liu, L., 2020

https://scholar.google.com/scholar?q=DQEAF:+Malware+Adversarial+Examples+Generation+with+Reinforcement+Learning

16. MalGAN: Generating Adversarial Malware Examples — Hu, W., Tan, Y., 2017

https://scholar.google.com/scholar?q=MalGAN:+Generating+Adversarial+Malware+Examples

17. IDSGAN: Generative Adversarial Networks for Attack Generation against Intrusion Detection — Lin, Z., Shi, Y., Xue, Z., 2022

https://scholar.google.com/scholar?q=IDSGAN:+Generative+Adversarial+Networks+for+Attack+Generation+against+Intrusion+Detection

18. Synthetic Data Generation for Cybersecurity: Challenges and Opportunities — Ring, M., Wunderlich, S., Scheuring, D., Landes, D., Hotho, A., 2019

https://scholar.google.com/scholar?q=Synthetic+Data+Generation+for+Cybersecurity:+Challenges+and+Opportunities

19. Generative Adversarial Networks for Black-Box API Attack on Web Application Firewalls — Duy, P.T., Khoa, N.M., Hoa, N.T., et al., 2022

https://scholar.google.com/scholar?q=Generative+Adversarial+Networks+for+Black-Box+API+Attack+on+Web+Application+Firewalls

20. Evading Anti-Malware Engines With Deep Reinforcement Learning — Fang, Z. et al., 2019

https://scholar.google.com/scholar?q=Evading+Anti-Malware+Engines+With+Deep+Reinforcement+Learning

21. Do LLMs Dream of Malware? Assessing the Practical Capability of LLMs to Generate Functional Malware — Pa Pa, Y. M. et al., 2023

https://scholar.google.com/scholar?q=Do+LLMs+Dream+of+Malware?+Assessing+the+Practical+Capability+of+LLMs+to+Generate+Functional+Malware

22. MITRE ATT&CK: Design and Philosophy — Strom, B. et al., 2018

https://scholar.google.com/scholar?q=MITRE+ATT&CK:+Design+and+Philosophy

23. LLM4Fuzz: Guided Fuzzing of Smart Contracts with Large Language Models — Various, 2024

https://scholar.google.com/scholar?q=LLM4Fuzz:+Guided+Fuzzing+of+Smart+Contracts+with+Large+Language+Models

24. Beyond the sandbox: Leveraging symbolic execution for evasive malware classification — approximate, ~2024, 2024

https://scholar.google.com/scholar?q=Beyond+the+sandbox:+Leveraging+symbolic+execution+for+evasive+malware+classification

25. Unveiling the dynamic landscape of malware sandboxing: A comprehensive review — approximate, ~2024, 2024

https://scholar.google.com/scholar?q=Unveiling+the+dynamic+landscape+of+malware+sandboxing:+A+comprehensive+review

26. TAMAS: Benchmarking Adversarial Risks in Multi-Agent LLM Systems — approximate, ~2024-2025, 2025

https://scholar.google.com/scholar?q=TAMAS:+Benchmarking+Adversarial+Risks+in+Multi-Agent+LLM+Systems

27. Red-teaming LLM multi-agent systems via communication attacks — approximate, ~2024-2025, 2025

https://scholar.google.com/scholar?q=Red-teaming+LLM+multi-agent+systems+via+communication+attacks

28. Assessing LLMs in malicious code deobfuscation of real-world malware campaigns — approximate, ~2024-2025, 2024

https://scholar.google.com/scholar?q=Assessing+LLMs+in+malicious+code+deobfuscation+of+real-world+malware+campaigns

29. Certifying accuracy, privacy, and robustness of ML-based malware detection — approximate, ~2024-2025, 2025

https://scholar.google.com/scholar?q=Certifying+accuracy,+privacy,+and+robustness+of+ML-based+malware+detection

30. AI Post Transformers: Petri: Accelerating AI Safety Auditing — Hal Turing & Dr. Ada Shannon, Fri,

https://podcasters.spotify.com/pod/show/12146088098/episodes/Petri-Accelerating-AI-Safety-Auditing-e39boei

31. AI Post Transformers: Bloom: an open source tool for automated behavioral evaluations — Hal Turing & Dr. Ada Shannon, Tue,

https://podcasters.spotify.com/pod/show/12146088098/episodes/Bloom-an-open-source-tool-for-automated-behavioral-evaluations-e3fi1ge

32. AI Post Transformers: Agentic Reasoning for Large Language Models: A Comprehensive Roadmap — Hal Turing & Dr. Ada Shannon, Sat,

https://podcasters.spotify.com/pod/show/12146088098/episodes/Agentic-Reasoning-for-Large-Language-Models-A-Comprehensive-Roadmap-e3e43ru

Interactive Visualization: MalGEN: Multi-Agent AI for Red Teaming Malware