Author : Miss. Sayali Sanjeev Satpute and Prof. Santosh Biradar

Active authentication is that the drawback of unceasingly substantiate the identity of an individual supported behavioral aspects of their interaction with a machine. During this paper, we tend to collect and analyze behavioral statistics knowledge from ten subjects, every victimization their personal android mobile device for an amount of a minimum of thirty days. This knowledge set is novel within the context of active authentication due to its size, duration, range of modalities, and absence of restrictions on tracked activity. The land collocation of the subjects in the investigation is illustrative of an enormous shut world condition like an enterprise any place the unapproved client of a gadget is probably going to be an insider risk: originating from inside the association. we consider four biometric modalities:

1. Text typed via soft keyboard;
2. Frequent calling and Contacts
3. Routine location
4. Frequent Network Or Data Connection

We implement and check a classifier for every modality and organize the classifiers as parallel binary decision fusion architecture. we are able to characterize the performance of the system with regard to intruder detection time and to quantify the contribution of every modality to the performance.

Keywords: Active authentication; application usage patterns; behavioral biometrics; decision fusion; GPS location; insider threat; intrusion detection; multimodal biometric systems; stylometry; web browsing behavior.