In this paper, we apply the uncomplicated decision tree data mining algorithm to find association rules about pornographic and medical web pages. On the basis of these association rules, we propose a systematized method of filtering pornographic websites with the following major superiorities: 1) Check only contexts of web pages without scanning pictures to avoid the low operating efficiency in analyzing photographs. Moreover, the error rate is lowered and the accuracy of filtering is enhanced simultaneously. 2) While filtering the pornographic web pages accurately, the misjudgments of identifying medical web pages as pornographic ones will be reduced effectively. 3) A re-learning mechanism is designed to improve our filtering method incrementally. Therefore, the revision information learned from the misjudged web pages can incrementally give feedback to our method and improve its effectiveness. The experimental results showed that each efficacy assessment indexes reached a satisfactory value. Therefore, we can conclude that the proposed method is possessed of outstanding performance and effectivity.
International Journal of Network Security, 19(5), 834-845.