The misuse of company devices is a widespread issue. In the police force however its far more severe. Police officers not only use their government assigned devices to spend time on Facebook while they should be on patrol. They also use these devices to contact criminals and request “favors” from them. Reducing the misuse of these devices is a central part in the fight against corruption and the misuse of government funds.
Since communicating with suspects and criminals is part of a police officer’s job, it is not practical to use content filters or firewalls. Using such means of prevention may interfere with the officer’s duties and compromise investigations.
Monitoring the officer’s device in real time and detecting when a misuse is about to occur is the preferred solution. However, a single device can collect a wide range of data such as GPS location, messages, calls, IMs, pictures, posts and more which need to be processed in real time.
Furthermore data describing what misuse constitutes is not available due to the nature of the subject matter involved.
To solve this issue, we took inspiration from the human neocortex and built a novel anomaly detection algorithm based around HTMs (hierarchical temporal memory cells). The algorithm takes in all the information collected by the officer’s device and then tries to determine if the user is exhibiting an unusual behavior. Then a signal is issued to another officer telling them to investigate the issue. Since HTMs learn in real time they don’t require any data to get started like most conventional algorithms. This makes them uniquely suited for this task.