Measure a traffic configuration before you trust it
Load packet captures from a circumvention client. See the traffic shape a passive observer sees, how far each session sits from ordinary internet traffic, and how your configurations rank under the PRAXIS score.
Sessions
Name each capture after the configuration it was recorded with. Captures with the same name are averaged as one configuration. Clear the checkbox for sessions that did not complete.
| File | Configuration | Packets | Duration (s) | Bursts | Benign distance | Size (bytes) | Completed | Remove |
|---|
Traffic shape
Each line is one packet’s payload. Above the axis is sent by the client, below is received. A burst is a run of packets in one direction.
| Feature | This session | Study, mux off | Study, mux c2 |
|---|
First 20 packets
The observer reads these signed payload lengths directly. Zero-payload packets, such as handshake and acknowledgment packets, show as short grey marks.
Rank configurations
The PRAXIS score combines detector confidence, distance from benign traffic, overhead and a completion penalty. Lower is better. How the score works
| Rank | Configuration | Sessions | Detector probability | Completed | Benign distance | Overhead | Score | Score parts |
|---|
What this does not do. It does not run the study’s detectors, which are not public. Enter a mean probability from your own detector to include that term. A low score ranks configurations; it does not show that a configuration evades detection. Confirm the winner on new sessions and with independent detectors, as the framework describes.