Online Binary Models are Promising for Distinguishing Temporally Consistent Computer Usage Profiles

This paper investigates whether computer usage profiles comprised of process-, network-, mouse-, and keystroke-related events are unique consistent over time in a naturalistic setting, discussing challenges opportunities using such applications continuous authentication. We collected ecologically-valid from 31 MS Windows 10 users 8 weeks submitted this data to comprehensive machine learning analysis involving diverse set online offline classifiers. found that: (i) were mostly the 8-week collection period, with most (83.9%) repeating habits on daily basis; (ii) profiling has potential uniquely characterize (with maximum F-score 99.90%); (iii) network-related relevant features accurately recognize (95.69% top distinguishing network-related); (iv) binary models well-suited for profile recognition, better results achieved setting compared (maximum 99.90% vs. 95.50%).