A keystroke dynamics based system for user identification

On-the-fly keyboard user authorization is certainly an interesting option for standard security procedures for high-security computer systems. Unauthorized access to logged-in workstation could threaten the security of data and systems. Conventional authorization methods (passwords, fingerprint scans) verify user identity only during logging process, leaving system vulnerable to user replacement afterwards. Procedures overcoming this peril are often invasive (constant visual monitoring) or uncomfortable (frequent identity verification with user cooperation). A possible solution for constant authorization without these drawbacks is to identify the keyboard user while typing. The approach we propose is based on two extracted keystrokes features: 'flight' and 'dwell'. We have tested the suggested solutions on individual group resembling a medium-sized company. The obtained results are promising.