We deployed Darktrace for one of the biggest telecommunications companies in Latin America. It is deployed on-premise, but it is more like a service because we don't care about the appliances. Even though it works with appliances, it is more related to the services to the connections that the solution can handle. Because of that, it is on-premise, but it also has a component with sensors that works for remote instances, almost like a cloud solution.
Some of the clients, especially in the security area, think that this appliance will replace a firewall or a prevention system solution, but it doesn't replace them. It actually complements them because the firewall decides to allow or deny a connection, and a prevention system is designed to avoid any type of risks to the connection or intrusion on the network. Darktrace allows you to find the unknown threats inside the network and identify them by using some artificial intelligence. It can do all the tracking inside or outside the network.
It is connected directly to the core switch, and in the first stage, it probably takes about a month to learn the behavior of the network and the users. With that, it starts to know what type of information is correct inside the network, and what type of information probably would be a risky connection or risky data moving from one site to another. It then starts doing the alerting. After the first stage or the learning stage is complete, we can find the size of the network. The second stage is the use of a different model inside the solution called Antigena. It works like the antibodies inside our body. Once it detects something that is wrong inside the network, it not only does the alerting but also takes the decision to block that type of connection in order to avoid any information leak or any possible risky connection. If somebody is doing some data mining, it disables connection to the engine that is doing the data mining.