Browser Engine Benchmark

A toolkit for testing browser automation engines against modern web protection systems. It checks how well each engine can bypass bot detection and measures their speed, resource usage, and resistance to fingerprinting.

Modern web applications use advanced bot detection like Cloudflare, DataDome, and Imperva to block automated access. This benchmark suite shows how different browser automation engines handle these defenses:

• Bypass Success Rate: Effectiveness against major protection systems
• Performance Metrics: Memory usage, CPU consumption, and page load times
• Fingerprinting Resistance: reCAPTCHA scores and CreepJS trust ratings
• Network Analysis: IP detection (proxy validation) and WebRTC leak testing

• Cloudflare
• DataDome
• Amazon
• Google Search
• Ticketmaster (Imperva)
• More systems coming soon

• Playwright - Microsoft's automation framework (Chrome, Firefox, Safari)
• Camoufox - Playwright-based
• Patchright - Playwright-based
• Playwright Stealth - Playwright-based
• Selenium - Open-source browser automation framework (apparently deprecated, so it is tested without proxies)
• NoDriver - Open-source browser automation framework (supports only SOCKS5 proxies)
• More engines coming soon. What engine should I add next?

• Automated report generation with visualizations
• Performance profiling and resource usage tracking
• Exportable results in JSON and Markdown formats

Using a clean proxy is essential for accurate benchmark results.

• IP Reputation: Your home/datacenter IP may already be flagged by protection systems from previous automation attempts, browser extensions, or security software
• Clean Testing Environment: A fresh proxy IP ensures you're testing the browser engine's capabilities, not your IP's reputation
• Rate Limiting: Repeated tests from the same IP can trigger rate limiting, affecting bypass success rates

This benchmark provides detailed comparative analysis. Here's an excerpt from a recent test run (more in results/example):
Real IP in this example - 169.150.201.11
Proxy IP in this example is different for each engine

If the CPU usage is 0 - failed to measure or it really is 0 for CDP sessions. The problem is known and will be fixed.

This Score is taken by solving the reCAPTCHA v3 on your browser. The Score shows if Google considers you as HUMAN or BOT. 1.0 is very likely a good interaction, 0.0 is very likely a bot With low score values (< 0.3) you'll get a slow reCAPTCHA 2, it would be hard to solve it. And vise versa, with score >= 0.7 it will be much easier.

Applicapable only with proxy. If the WebRTC IP is different from your real IP - no leakage

Applicapable only with proxy. If the IP is your proxy's IP - good, your real IP - bad.

1. Clone the repository

   git clone https://github.com/techinz/browsers-benchmark.git cd browsers-benchmark

2. Set up Python environment

   python -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate pip install -r requirements.txt

3. Install browser engines

   Playwright

   playwright install # On Linux also run: playwright install-deps

   Camoufox

   # Windows camoufox fetch # Linux python -m camoufox fetch sudo apt install -y libgtk-3-0 libx11-xcb1 libasound2

   Patchright

   patchright install chromium

4. Configure settings

   cp .env.example .env # Edit .env with your proxy settings if needed

5. Configure proxies

   1. Create a file named proxies.txt in the documents directory.
   2. Add your proxy URLs in format http://username:password@proxy_host:port or http://proxy_host:port.
      ❗️ IMPORTANT (1): Number of proxies has to be not less than number of engines you want to test.
      ❗️ IMPORTANT (2): Some engines support different proxy protocols - for example, Playwright supports only HTTP and HTTPS, but NoDriver supports only SOCKS5.
      This implies that you have to add multiple proxy protocols to the proxies.txt file or exclude some engines from the test.
      At the moment you need all HTTP/HTTPS proxies and at least 1 SOCKS5 for NoDriver. Also, the benchmark will show you what proxy protocols are missing.
      ❗️ IMPORTANT (3): Selenium won't use any proxies.

   Example proxies.txt content (each line is a separate proxy):

   http://proxy1.example.com:8080 http://proxy2.example.com:8080 http://username:[email protected]:8080 http://username:[email protected]:8080 socks5://username:[email protected]:8080

6. Run benchmark

The benchmark generates reports in the results/ directory:

• summary.md - Human-readable markdown report
• benchmark_results_*.json - Raw data for further analysis
• media/ - Generated visualizations and screenshots
  • bypass_dashboard.png - Multi-metric dashboard
  • recaptcha_scores.png - reCAPTCHA performance chart
  • creepjs_scores.png - Fingerprinting resistance analysis
  • screenshots - Screenshots of all tested targets

The codebase follows a modular architecture for extensibility:

1. Modify config/benchmark_targets.py to add custom test targets:

   Target( name="custom_site", url="https://example.com", check_function="check_custom_bypass", description="Custom site protection test" )

2. Create a check function for the target in utils/targets/check_bypass, for example in a file named custom_bypass.py:

   from engines.base import BrowserEngine async def check_custom_bypass(engine: BrowserEngine) -> bool: element_found, element_html = await engine.locator('//div[@class="captcha"]') return not element_found # no captcha found - success!

3. Add it to the checkers mapping in config/benchmark_targets.py's BypassTargetsSettings:

   checkers: Dict[str, Callable] = Field( default_factory=lambda: { "check_cloudflare_bypass": check_cloudflare_bypass, "check_datadome_bypass": check_datadome_bypass, ... "check_custom_bypass": check_custom_bypass, } )

1. Extend the BrowserEngine base class:

   class CustomEngine(BrowserEngine): async def start(self) -> None: # Initialize browser async def navigate(self, url: str) -> Dict[str, Any]: # Navigation logic

   Or, if Playwright-based, extend PlaywrightBase base class:

   class CustomPlaywrightBasedEngine(PlaywrightBase): ...

   Or, if Selenium-based, extend SeleniumBase base class:

   class CustomSeleniumBasedEngine(SeleniumBase): ...

2. Add it to the engines mapping in config/engines.py's EnginesSettings:

   base_engines = [ { "class": PlaywrightEngine, "params": {"headless": True, "name": "playwright-chrome_headless", "browser_type": "chromium"} }, ... { "class": CustomEngine, "params": {"headless": True, "name": "custom_engine", "browser_type": "chromium"} } ]

Common Issues:

• Detection failures: Verify proxy configuration and target accessibility

Contributions are welcome! Areas where help is needed:

• New Protection Systems: Add support for additional bot detection services
• Browser Engines: Implement support for new automation frameworks (e.g. Selenium-based)
• Analysis Tools: Enhance reporting and visualization

This project is licensed under the MIT License - see the LICENSE file for details.

This tool is designed for educational and research purposes. Users are responsible for ensuring compliance with website terms of service and applicable laws. The authors and contributors do not encourage or endorse any malicious use of this software.