📱 Smart DPI-aware client app — automatic protocol selection + peer intelligence

[sepehrsadri]

The Problem

Currently, our smart-sub worker delivers configs and users manually try different protocols until one works. There is zero intelligence about what's blocked and what isn't.

Vision: Smart DPI-Aware Client

A client app that:

1. Detects the user's ISP and current DPI rules (passive fingerprinting)
2. Selects the best protocol automatically (based on success history + peer data)
3. Adapts in real-time when a protocol gets blocked mid-session
4. Reports DPI behavior back to a shared intelligence network

┌─────────────────────────────────┐
│        Smart VPN Client         │
├─────────────────────────────────┤
│  DPI Detection Engine           │
│  ├── ISP fingerprinting         │
│  ├── Protocol probe (background)│
│  └── Latency measurement        │
├─────────────────────────────────┤
│  Protocol Selection Engine      │
│  ├── Success history (local)    │
│  ├── Peer intelligence (remote) │
│  └── Weighted random selection  │
├─────────────────────────────────┤
│  Connection Manager             │
│  ├── Auto-fallback chain        │
│  ├── Parallel connection racing │
│  └── Session persistence        │
└─────────────────────────────────┘

Research Tasks

• Research libbox (sing-box library) API — what can be controlled programmatically?
• Prototype Flutter app using libbox that connects to one VLESS server
• Design protocol selection algorithm (multi-armed bandit? weighted scoring?)
• Build ISP detection module (what ISP is the user on? Irancell/MCI/Rightel/fixed?)
• Design peer intelligence API (users anonymously report working protocols per ISP)
• Evaluate TestFlight for iOS distribution inside Iran (10k user limit, 90-day builds)
• Research APK distribution channels for Iran (Myket, Bazaar, Telegram direct)

Key References

• Hiddify App Source — Flutter + libbox
• sing-box libbox
• RESEARCH-TOPICS.md — Topics 1 & 2