DETECTING WEB-BASED BOTNETS USING A WEB PROXY AND A CONVOLUTIONAL NEURAL NETWORK
Botnets are increasingly becoming the most dangerous threats in the field of network security, and many different approaches to detecting attacks from botnets have been studied. Whatever approach is used, the evolution of the botnet's nature and the set of defined rules for detecting botnets ca...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Dalat University
2020-09-01
|
| Series: | Tạp chí Khoa học Đại học Đà Lạt |
| Subjects: | |
| Online Access: | http://tckh.dlu.edu.vn/index.php/tckhdhdl/article/view/652 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Botnets are increasingly becoming the most dangerous threats in the field of network security, and many different approaches to detecting attacks from botnets have been studied. Whatever approach is used, the evolution of the botnet's nature and the set of defined rules for detecting botnets can affect the performance of botnet detection systems. In this paper, we propose a general family of architectures that uses a convolutional neural network group to transform the raw characteristics provided by network flow recording and analysis tools into higher-level features, then conducts a (binary) class to assess whether a flow corresponds to a botnet attack. We experimented on the CTU-13 dataset using different configurations of the convolutional neural network to evaluate the potential of deep learning on the botnet detection problem. In particular, we propose a botnet detection system that uses a web proxy. This technique can be helpful in implementing a low-cost, but highly effective botnet detection system. |
|---|---|
| ISSN: | 0866-787X 0866-787X |