An Effective Method to Generate Simulated Attack Data Based on Generative Adversarial Nets
Huihui Xie, Kun Lv, Changzhen Hu · 2018
In practice, there are few available attack dataset. Although there are many methods that can be used to simulate cyberattacks for attack data, such as using specific tools, writing scripts to simulate the attack scenes, etc.. The disadvantages of those methods are also obvious. Tools developers and script authors need to know professional network security knowledge. As tools are implemented in different ways, users also need to have some expertise. What's more, it may take a long time to generate a large amount of attack data. In this paper, we present some of the existing network attack tools and proposed a method to generate attack data based on generative adversarial network. Using our method you do not need to have a professional network security knowledge, only use some basic network attack data one can generate a large number of attack data in a very short period of time. As network malicious activities become increasingly complex and diverse, network security analysts face serious challenges. Our method also can generate mixed features attack data by setting training data with different attack types. It has high performance. To test the performance of our method, we did a test and found that it took only 160 seconds to generate a million connection records in a PC with 3.7GHz, 4 core CPU and 8G memory.