Skip to main content

About the project

just realized i missed the deadline. not sure how to remove this entry, but the repo has been moved to private, honeypot is torn down and nothing sensitive is in the screenshots. terminal screenshot was the honeypot with made up data. Thanks for letting me particpate! till next time :)

Honeypot-as-Code with an open-weight LLM playing the server.

Judges: start with docs/JUDGES.md

Try out the honeypot! ssh -o PubkeyAuthentication=no -o PreferredAuthentications=password [email protected]

all returns from commands are fabricated/made up.
read up on how this part works here: https://github.com/dacoty/mirage/blob/main/docs/HOW_IT_WORKS.md


AI usage statement: https://github.com/dacoty/mirage#ai-usage-statement

Claude file: https://github.com/dacoty/mirage/blob/main/CLAUDE.md
Spec: https://github.com/dacoty/mirage/blob/main/SPEC.md

Built with

pythonclaudegroq

AI tools

Coded withClaude CodeClaude Sonnet 5.5Qwen3.8 27B

Prize Categories Entered

  • DigitalOcean

    Best Use of DigitalOcean

    Submitted

    Mirage runs entirely on DigitalOcean, provisioned with Terraform. One terraform apply creates a single Ubuntu Droplet (about $0.021/hour) running the honeypot, TimescaleDB and Grafana in Docker, bootstrapped by cloud-init. A Cloud Firewall opens port 22 to the world as the decoy. Hourly billing and terraform destroy let me run a real internet-facing honeypot for cents and tear it all down afterward.

  • DigitalOcean

    Best Open-Source AI Project

    Submitted

    Mirage runs entirely on DigitalOcean, provisioned with Terraform. One terraform apply creates a single Ubuntu Droplet (about $0.021/hour) running the honeypot, TimescaleDB and Grafana in Docker, bootstrapped by cloud-init. A Cloud Firewall opens port 22 to the world as the decoy. Hourly billing and terraform destroy let me run a real internet-facing honeypot for cents and tear it all down afterward.