Summary of an article published on The Web Scraping Club on 2026-03-22 (free to read). Read the full article
Live data on modern sites often travels over WebSockets, so connecting straight to the channel looks like the obvious shortcut for scraping. It is possible, but it comes with its own obstacles. Clients are fewer than for HTTP, documentation is thin, proxy support is uneven, and there is no request-response model, only a stream of events to handle in real time. On top of that, servers can detect bots in ways HTTP never allowed, because the connection is stateful and persistent.
I walk through the WebSocket-specific detection techniques you meet when targeting a WS server from a Python or Node.js script: handshake validation, honeypot events and channels, connection lifecycle patterns, and binary payloads, using TikTok LIVE as the example. Each comes with practical tips. I then cover what WAFs and other anti-bot systems do with WebSockets, which matters also if you use Playwright or Selenium: TLS fingerprinting, continuous device fingerprinting, and real-time behavior monitoring.
What’s inside
- Which headers a browser sends in the upgrade request beyond what the RFC requires, and why they feed fingerprinting.
- How honeypot events and hidden channels catch bots that react to every message or subscribe to channels they should not use.
- Which lifecycle signals give a client away, including connection churn per IP and the unnaturally stable latency of data center scripts.
- Why binary, compressed or encrypted payloads are costly to reverse-engineer, and the alternative data source I suggest for TikTok LIVE.
In the full article: The header example for the handshake, the full tip lists for each technique, the DevTools screenshots of TikTok LIVE binary frames and its HTTP API calls, and the links to related TWSC posts.