This paper presents a scheme to support curtailing illegal activities that are carried out with the help of computers. The paper focuses on determining criminal character of a user by analyzing user’s interactions with the computer at the operating system level. Doing this at the operating system level gives an advantage of catching all interactions of the user with the computer. User interaction information is obtained during the system use and this information is classified using neural network. Neural network does the processing to obtain the criminal character of the user. A sample test was conducted on 200 different users (50 criminal users and 150 normal users). The results reported show that the proposed system is practical and accurate.