With the development of wireless communication, cyber physical system (CPS) technologies are being applied to various fields, and people's daily lives are more dependent on CPS. As CPS brings convenience to people's lives, danger also arises. The most serious of these is attacks on CPS. Attackers obtain information without the user's permission. The main transmission medium used by attackers is the botnet. The domain generation algorithm is mainly used in botnets. This algorithm generates and registers a large number of domain names in a very short time for CPS, and then binds the IP address of the botnet controller. Due to the development of domain generation methods, the detection of such domains is crucial for security in CPS but has stagnated. To end the situation, this article proposes a domain name detection system to solve this security issue in CPS. In the system, a deep learning powered adversarial sample attacks approach is embedded to improve its performance. Through experiments, the proposed system achieves better performance in malicious domain name recognition.

中文翻译：

---

用于网络物理系统中 DGA 域名安全检测的深度学习驱动的对抗样本攻击方法


随着无线通信的发展，信息物理系统（CPS）技术正在应用于各个领域，人们的日常生活更加依赖于CPS。 CPS在给人们生活带来便利的同时，也带来了危险。其中最严重的是对 CPS 的攻击。攻击者未经用户许可获取信息。攻击者使用的主要传输媒介是僵尸网络。域名生成算法主要用于僵尸网络。该算法在很短的时间内为CPS生成并注册大量域名，然后绑定僵尸网络控制器的IP地址。由于域生成方法的发展，此类域的检测对于 CPS 的安全至关重要，但目前已经停滞不前。为了结束这种情况，本文提出了一种域名检测系统来解决CPS中的这一安全问题。在该系统中，嵌入了深度学习驱动的对抗性样本攻击方法，以提高其性能。通过实验，该系统在恶意域名识别方面取得了较好的性能。