An online neural network system for computer access security

A method for identifying computer users based on the individual typing techniques of the users is presented. The identification system is a pattern classification system based on a simulation of an artificial neural network. The user types a known sequence of characters, and the intercharacter times represent a pattern vector to be classified. This vector is presented to the classification system, and the pattern is assigned to a predefined class, thus identifying the user. The major work is divided into two phases: the investigation phase and the implementation phase. Experimental results are discussed, followed by a description of a real-time implementation of this system, using a personal computer, known as the OnLine User Identification System. In an operational trial, the system correctly identified users 97.8% of the time. This intelligent system can be used, in addition to the traditional user name and password procedures, to improve computer security in a cost-effective manner. >