🧠 Geneva — automated DPI-evasion discovery via genetic algorithms

[sepehrsadri]

Background

Geneva (Genetic Evasion) is a research tool from University of Maryland that uses genetic algorithms to automatically discover censorship evasion strategies. Instead of manually crafting bypass techniques, Geneva evolves them.

Why This Matters for Iran

Iran's DPI is actively updated. Manual protocol tuning is a cat-and-mouse game. Geneva can:

• Automatically discover packet manipulation strategies that bypass Iran's specific DPI boxes
• Find evasion techniques humans wouldn't think of
• Re-run periodically as Iran updates their filters
• Generate strategies that can be encoded into our client app's protocol selection

Research Tasks

• Document discovered strategies per ISP (Irancell, MCI, Rightel)
• Evaluate if discovered strategies can be integrated into sing-box/xray transport
• Test discovered strategies' durability (how long before Iran patches?)
• Create an automated pipeline: Geneva discovers → strategy deployed → tested → rotated

Setup Requirements

Iran-side machine (client)  ←→  Our VPN server (server-side)
         ↕
    Geneva client              Geneva server
    (evolves strategies)       (evaluates fitness)

• Server-side: any of our 4 servers
• Geneva is Python-based, runs on Linux

References

• Geneva Project
• Geneva GitHub
• Geneva Paper
• GFW Report — similar DPI analysis for China
• RESEARCH-TOPICS.md — Topic 3d

---

Note on in-country testing. Measurement from inside a censored network is valuable, but this tracker is public: replying here to volunteer means self-identifying with a GitHub account. Please don't. Results are welcome via any channel you're comfortable with, and RIPE Atlas (already wired up in ripe-atlas-check.yml) gives a measurement path that puts no individual at risk.