VantaBlock

A unified behavioural logging system that collects and analyzes user interactions (both human and bot) to detect automated traffic. Built for the HackByte 4.0 hackathon.

What Is Implemented

• Unified event schema for both human (frontend/logger.js) and bot (bot/bot_script.py).
• Backend endpoint POST /api/logs that saves one JSON file per session in logs/.
• Session file naming convention:
  • human_YYYYMMDD_HHMMSS_uniqueID.json
  • agent_YYYYMMDD_HHMMSS_uniqueID.json
• Feature extraction pipeline: extract_features.py writes a CSV table for ML training.

MockShop Behavioural Logging (Unified Schema)

This workspace collects comparable behavioural data for human and agent sessions.

Run

1) Start backend (serves pages + log API)

cd backend
python main.py

Backend runs on http://localhost:8001 and serves frontend at http://localhost:8001/demo/index.html.

2) Collect human session

Open in browser:

• http://localhost:8001/demo/index.html

Interact naturally across pages; logger sends full session snapshots every 30 seconds and on page unload.

3) Collect bot session

cd bot
python bot_script.py

The bot writes a local session file to logs/ and also POSTs to /api/logs.

4) Inspect stored sessions

• http://localhost:8001/sessions?session_type=human
• http://localhost:8001/sessions?session_type=agent
• http://localhost:8001/session/<session_id>

5) Extract ML features

From project root:

python extract_features.py --logs-dir logs --output features.csv

Session JSON Shape

{
  "session_id": "human_20260404_153022_42ab9c",
  "session_type": "human",
  "start_time_ms": 0,
  "end_time_ms": 123456,
  "user_agent": "Mozilla/5.0 ...",
  "viewport": { "width": 1280, "height": 800 },
  "events": [
    {
      "event_id": 1,
      "timestamp_ms": 0,
      "type": "navigation",
      "data": { "url": "...", "referrer": "" }
    }
  ]
}

Built for HackByte 4.0. Uses Playwright for browser automation and scikit‑learn for behavioural classification.