Melisseus

Large Language Models as Defensive Honeypots

ProjectID

NGI-SARGASSO:Melisseus

Acronym

Melisseus

Additional Info

To research, develop, and apply LLMs to generate honeypots designed to deceive human attackers. They aim to increase variance in response types, improve technical detail, and precisely replicate different target systems. We expect the LLM's generated output to be more effective at luring or confounding attackers into perceiving honeypots as genuine production systems. We plan to use LLMs to create honeypots tailored to popular protocols like SSH, TELNET, SMTP, POP3, MySQL, MsSQL, and HTTP.

Enduser Relevance

Melisseus develops advanced AI-powered honeypots that simulate human-like interactions to deceive attackers. It focuses on local, open-source LLMs for dynamic defenses, enhancing security by engaging and identifying threats more effectively.

Contact

sebastian.garcia@agents.fel.cvut.cz

Endorsements

Not available yet

Disclaimer

Not available yet

Country:  Czech Republic

Status: Early research demo

Category: Decentralized solutions (including blockchain and distributed ledger technologies)

check other similar innovations
Skip to content