PyTorrent: Asyncio BitTorrent Client

A lightweight, asynchronous BitTorrent client built from scratch in Python 3.6 using asyncio. Designed to demonstrate networking protocols, binary data manipulation, and concurrent I/O.

😋 USAGE

Download and save the torrent file in the same directory as main.py.

python main.py <torrent_file>

🚀 Features (Current Focus)

• Bencoding Parser/Encoder: Full support for decoding .torrent files and encoding info dictionaries for hashing.
• Tracker Communication: HTTP/UDP support to retrieve peer lists.
• Async Peer Protocol: Event-driven TCP connections using asyncio.
• Piece Management: Sequential download strategy with block assembly.
• Concurrency: Simultaneous downloading from multiple peers.

🔮 Roadmap ([ADV] Features)

• Rarest-First Algorithm: Optimized piece selection strategy.
• Pause/Resume: State persistence for long downloads.
• Seeding: Upload support for other peers.
• Streaming Mode: Prioritize sequential pieces for video playback.
• AI Optimization: Smart peer selection/scoring using simple heuristics or ML.
• Privacy Mode: SOCKS5 Proxy support to mask IP.
• Protocol Obfuscation: To bypass basic ISP throttling (MSE).
• Content Safety: ML-based classification of file metadata.
• Reinforcement Learning: Algorithm for peer selection.

😶‍🌫️Further Details on [ADV]:

• Reinforcement Learning:

  • Currently, clients mostly assume "Tit-for-Tat" (if you give me speed, I give you speed).
  • AI agent could learn which peers are "bursty" or "reliable" over time, predicting who will unchoke you faster than standard algorithms.
  • We will thus use Multi-Armed Bandit (Reinforcement Learning) algorithm for peer selection.

• Masking IP / VPN:

  • You cannot "mask" your IP in P2P without a relay. If you don't send your IP, the peer can't send you data.
  • SOCKS5 Proxy Support. We can add a feature to route all traffic through a proxy (like Tor or a commercial VPN proxy). This is how standard clients hide the user's IP.

• ML for Legal/Illegal Detection:

  • Challenge: You only see binary chunks, not the image/video content, until it's done.
  • NLP on Metadata.
  • Flag suspicious keywords or patterns often associated with malware or illegal content.

• Avoid ISP Blocking

  • ISPs block Torrenting by "Deep Packet Inspection" (DPI)—they see the handshake pattern.
  • Solution: Message Stream Encryption (MSE/PE).
  • Wraps the BitTorrent header in an encrypted stream (RC4) so it looks like random noise to the ISP, rather than a torrent.

🛠️ Technology Stack

• Language: Python 3.6+
• Concurrency: asyncio (Event Loop)
• Networking: aiohttp (Tracker HTTP), asyncio streams (TCP Peers)
• Data: struct (Binary packing), hashlib (SHA1 integrity)

📂 Architecture

The project is modularized to separate protocol logic from network management:

• utils/bencoder.py: Handles the raw data serialization.
• core/tracker.py: Manages the initial discovery of peers.
• core/peer.py: The state machine for individual peer connections (Handshake -> Choke/Unchoke -> Request).
• network/download_manager.py: The orchestrator that assigns work to peers.