Most security games put you in the attacker's chair. This one puts you in ours. Traffic arrives, some of it hostile and most of it paying customers, and you decide which filters run and how hard they push.
Five shifts. No signup. Nothing you do leaves your browser.
Start your shift How the real firewall worksEvery demo of a security filter shows it catching an attack. That is the easy half, and it is the half that makes a product look good. The hard half is the customer who writes "please disregard the previous instructions, the delivery date has changed" and gets stopped by a filter doing exactly what it was built to do. That message scores 0.55 on the pattern filter and 0.76 on the classifier. It is a real sentence a real customer sends.
Turn everything on and set it to strict and you will catch nearly every attack in this game. You will also block people who were trying to pay you. Somewhere between those two failures is the setting a real team has to choose, and choosing it is the job this console hands you.
The same detection code as the rest of this site's demonstrations, running in your browser on the traffic in front of you. No model is judging you and nothing is scripted: arm a different set of layers and the same message scores the same way it always would, because the score belongs to the message and the layer, not to the game.
The target at the end of each shift is measured too. It is the best result achievable by playing that shift through every configuration the latency budget allows, which is why it is occasionally lower than you expect. Some shifts cannot be played cleanly, and a target written by hand would have hidden that.
No. It is a working reimplementation of the detection logic in the SecuritAI AI Firewall, not the production service, and each layer says where the two part company. The latency figures are illustrative, chosen to show the shape of the tradeoff rather than to report a measurement. What is real is the detection: the byte inspection, the pattern families, the intent matching and the scoring all run for real on the traffic you are judging.
If you would rather attack than defend, The Seventh Layer is the same engine from the other side: talk an assistant out of a key while these layers try to stop you. And the AI Firewall runs this pipeline in front of a real model in production.