| Summary: | 碩士 === 國立彰化師範大學 === 數學系所 === 101 === Since the development and wide scale adoption of internet technology, communications software such as Skype, Line, or WeChat has made interpersonal communication fast and convenient. However, because of the difficulty of confirming on-line identity and the ease with which account security can be violated, in recent years many researchers have published on the topic of keystroke dynamics in biometric authentication with an eye towards enhancing account security.
Despite the interest in it, keystroke dynamics used to identify passwords with fixed content and length do not apply to free-text recognition on the instant messaging software. In recent years, the biggest drawbacks related to research in free-text is that the training time requires a few months, a big problem for its practicality, and also that the system generates too many false alarms. Those issues for free-text instant messaging identification reduced user demand.
In order to solve the above mentioned problems, in this paper we propose a prediction mechanism for user's keystroke dynamics; each user only requires about 20 minutes of keystroke training to identify instant messages through this system. Furthermore we propose a Voting-Based Statistical classifier in order to improve the recognition accuracy of instant messages and prevent phishing messages. Finally, in this paper we propose using a keyword based identification method to reduce the problems with excessive false alarms. Experimental results show that the system proposed in this paper shows better result regarding training time and also the number of false alarms than other relevant published research. We also reached a comparable recognition accuracy.
|
|---|
