Abstract | ||
---|---|---|
Online phishing usually tricks victims by showing fake information which is similar to the legitimate one, so that the phishers could elevate their privileges. In order to guard users from fraudulent information and minimize the loss caused by visiting phishing websites, a variety of methods have been developed to filter out phishing websites. At present, there are several phishing detection methods continually being updated, but the experimental results of them are not enough satisfactory. To fill these gaps, an improved model based on attention mechanism bi-directional gated recurrent unit, named BiGRU-Attention model, will be introduced. The basic mechanism of this model is that it obtains the characters before and after a particular character through the BiGRU, and then calculates score for that character by the Attention. Since the final score depends on the composition of the input, the more similar between phishing and legitimate websites, the more difficult it is to be distinguished. By utilizing this model, most of the phishing URLs will be tested out. Also, an explanation of why phishing and legal websites can be distinguished will be given. Based on the experimental results, the BiGRU-Attention model achieves an accuracy of 99.55%, and the Fl-score is 99.54%. Besides, the effectiveness of deep neural network in anti-phishing application and cybersecurity will be demonstrated. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1007/978-3-030-41579-2_43 | INFORMATION AND COMMUNICATIONS SECURITY (ICICS 2019) |
Keywords | DocType | Volume |
Phishing detection, BiGRU-Attention model, Important characters, The difference between similar URLs | Conference | 11999 |
ISSN | Citations | PageRank |
0302-9743 | 0 | 0.34 |
References | Authors | |
0 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Lijuan Yuan | 1 | 0 | 0.34 |
Zhiyong Zeng | 2 | 0 | 0.34 |
Yikang Lu | 3 | 0 | 0.34 |
Xiaofeng Ou | 4 | 0 | 0.34 |
Tao Feng | 5 | 0 | 0.34 |