README.md

Fingerproxy

Inspired by gospider007/fp. Fingerproxy is an HTTPS reverse proxy. It creates JA3, JA4, Akamai HTTP2 fingerprints, and forwards to backend via HTTP request headers.

         HTTPS                            HTTP/1.1 or HTTP/2
Client ---------> Fingerproxy   -------------------------------------->  HTTP Backend
                 (listens :443)    | Request headers:               |   (127.0.0.1:80)
                                   |   X-JA3-Fingerprint: abcd...   |
                                   |   X-JA4-Fingerprint: t13d...   |
                                   |   X-HTTP2-Fingerprint: 3:100...|

Fingerprints can be used for bot detection, DDoS mitigation, client identification, etc. To use these fingerprints, just extract the HTTP request headers in your backend apps.

Fingerproxy is also a Go library, which allows users implementing their own fingerprinting algorithm.

Usage

Tip

Try fingerproxy in 1 minute:

1. Generate a self-signed certificate tls.crt and tls.key for testing.

   openssl req -x509 -newkey ec -pkeyopt ec_paramgen_curve:secp384r1 -days 3650 \
     -nodes -keyout tls.key -out tls.crt -subj "/CN=localhost" \
     -addext "subjectAltName=DNS:localhost,DNS:*.localhost,IP:127.0.0.1"

2. Download the fingerproxy binary and run.

   ./fingerproxy -listen-addr :8443 -forward-url https://httpbin.org

   Server listens on :8443, forwarding requests to httpbin.

3. We are ready to go. Send an HTTPS request to fingerproxy:

   curl "https://localhost:8443/anything?show_env=1" --insecure

   You will see fingerprints in request headers:

   {
     "headers": {
       # ...
       "X-Forwarded-Host": "localhost:8443",
       "X-Forwarded-Port": "443",
       "X-Forwarded-Proto": "https",
       "X-Http2-Fingerprint": "3:100;4:10485760;2:0|1048510465|0|m,s,a,p",
       "X-Ja3-Fingerprint": "0149f47eabf9a20d0893e2a44e5a6323",
       "X-Ja4-Fingerprint": "t13d3112h2_e8f1e7e78f70_6bebaf5329ac"
     },
     # ...
   }

For all available CLI options, see fingerproxy --help.

Production-Ready

The fingerproxy binary is production-ready. Subscan.io has 12 fingerproxy instances running in the production environment, which process almost 40,000,000 requests/day on average.

Unit tests, memory usage tests, E2E tests, and benchmarks have been implemented and run on GitHub Actions.

And of course, fingerproxy follows SemVer.

Kubernetes and Prometheus Integration

Kubernetes liveness probe support is available since v0.3.0. Example:

apiVersion: v1
kind: Pod
metadata:
  name: fingerproxy
spec:
  containers:
  - name: fingerproxy
    image: fingerproxy
    livenessProbe:
      httpGet:
        path: /
        port: 443
        scheme: https

Kubernetes probes use certain User-Agent such as kube-probe/1.26. Therefore, those requests with the specific user-agent header can be recognized as probing requests. Instead of forwarding to the backend app, fingerproxy will simply respond with an HTTP 200 by itself.

The default Prometheus metrics server listens on :9035. Once new requests come in, run curl http://localhost:9035/ to see avaialble metrics.

Implement Your Fingerprinting Algorithm

Check out the examples ja3-raw or my-fingerprint. No code fork needed.

Chrome JA3 Fingerprints Change Every Time

Yes, it is an known issue of the original JA3 implementation. See Google Chrome TLS extension permutation. Sorting the TLS extension is one method to avoid the affect of this feature. Here is an example ja3-sorted-extensions.

Use as a Library

Fingerproxy is degigned to be highly customizable. It is separated into serveral packages. You can find all packages in the pkg dir and use them to build your own fingerprinting server.

Here is an example echo-server. Instead of forwarding HTTP requests, it simply responds back to client with the fingerprints.

Similar Projects

• gospider007/fp

  Great implementation based on golang net stack, works nice with golang HTTP handler. Fingerproxy rewrites TLS ClientHello capturing according to it.

  Why I didn't use it?

  • The JA3 and JA4 implementations contain bugs. For example,
    • In fp.go, _ should be used instead of , as the separator of extensions and signature algorithms.
    • In fp.go, the protocol version in ClientHello handshake should be used for JA3.
    • In ja3.go, ; should be used instead of , as the parameters separator in HTTP2 settings frame.
    • Priority frame used in HTTP2 fingerprinting has not implemented yet.
  • KISS
    • JA4H can be calculated in backend applications. Do not do it in the reverse proxy.
    • Complex design, unused methods, and copy-pastes from unknown source.

• sleeyax/ja3rp

  Discontinued.

• h3adex/guardgress

  Relies on gospider007/fp.

• wwhtrbbtt/TrackMe

  An HTTPS echo server that responds the fingerprints.

  Why I didn't use it?

  • It is not designed for forwarding the requests.
  • Bugs too. In ja4.go, the padding extension (21) somehow has preserved. Therefore there would be two duplicated 21 extension which is incorrect.

Useful Websites

• https://browserleaks.com/tls

  JA3 and JA3 with sorted TLS extensions.

• https://scrapfly.io/web-scraping-tools/ja3-fingerprint

  JA3 and HTTP2 fingerprint. As per the comment from Scrapfly team member, Scrapfly uses an improved variant of JA3 implementation. It makes more sense in real life use cases, however, please notice that these JA3 results are not comparable with others. For more information, please read issue #14.

• https://tls.peet.ws/

  Public deployment of wwhtrbbtt/TrackMe; JA4 result might be incorrect (see above).

References

• JA3 fingerprint: https://github.com/salesforce/ja3
• JA4 fingerprint: https://github.com/FoxIO-LLC/ja4/blob/main/technical_details/JA4.md
• Akamai HTTP2 fingerprinting: https://www.blackhat.com/docs/eu-17/materials/eu-17-Shuster-Passive-Fingerprinting-Of-HTTP2-Clients-wp.pdf