TLS Fingerprints
What JA3 and JA4 identify, and what they cannot tell you.
A TLS fingerprint hashes the parts of a handshake a client picks for itself: versions, ciphers, extensions and their order. Those come from the TLS library, so the fingerprint follows the software, not the network.
That is the strength of the technique and the source of every way it gets misused. Many programs share a library.
3 rows.
| Scheme | Strength | Weakness |
|---|---|---|
| JA3 (client) | Identifies the TLS implementation and how it was configured, so it survives an infrastructure change. Malware that moves to new servers keeps the same JA3 until its TLS stack changes. | ORDER-SENSITIVE and therefore brittle. A library update, a different build, or GREASE values shifting positions changes the hash for entirely benign reasons. Two unrelated tools built on the same library also collide. |
| JA3S (server) | Pairing JA3 with JA3S is far stronger than either alone: it fingerprints the conversation rather than one end of it, and C2 servers are often more distinctive than their clients. | Much lower entropy than JA3, because a server picks one cipher rather than offering dozens. On its own it is close to useless. |
| JA4 | Deliberately fixes JA3's two worst properties: the lists are SORTED before hashing, so ordering churn and GREASE stop mattering, and the string is partly human-readable, so you can compare pieces rather than all-or-nothing. | Newer, so support and shared threat-intelligence coverage are thinner than JA3's. It is still a fingerprint of an implementation, not of an actor. |