الفهرس يوجد فقط 14 صفحة متاحة للعرض العام
 |  | from | 94 |  |  |
| | | from | 94 | | |
المستخلص
Biometric has been an increasing demand for many security systems to provide more accuracy of individual’s identification using computer. One of these identification methods is keystroke dynamic. A keystroke dynamic is a biometric measurement in terms of keystroke press duration and keystroke latencies. Keystroke dynamics application can be built using data mining techniques. Specially, keystroke biometric technology can provide more accurate and affect level of security of electronic commerce. To this end, many classification techniques have been proposed to achieve a high degree of identification accuracy. However, several problems are arisen such as similar user’s behavior and achieving high accuracy.
The main contributions of this thesis are twofold. The first is to adapt the data mining techniques to achieve high accurate-security without requiring any special hardware (using Keyboard only). To solve similar user problem, we classify users data based on a membership function into fuzzy sets. Next, we employed a sequence alignment algorithm (Needleman-Wunch (NM&W)) as a way of pattern discovery from user typing behavior.
The second contribution aims to utilize the ant colony optimization algorithm ACO which is introduced to solve computational problem and provide other alternative solutions. ACO algorithm in its original construct can’t handle all types of data. We proposed a model to reconstruct the ACO algorithm to handle the fuzzified keystroke sequences from one side. Additionally, we introduced a customized selection and pheromone update based on the NM&W alignment algorithm.
The conducted experiments proved that our proposed system improves the accuracy and precision by more than 30% and 40% respectively compared with traditional classification methods.