Department of Information Management, National Pingtung University, Pingtung 900, Taiwan.
Sensors (Basel). 2023 Jul 4;23(13):6139. doi: 10.3390/s23136139.
Distributed Denial of Service (DDoS) attacks pose a significant threat to internet and cloud security. Our study utilizes a Poisson distribution model to efficiently detect DDoS attacks with a computational complexity of O(). Unlike Machine Learning (ML)-based algorithms, our method only needs to set up one or more Poisson models for legitimate traffic based on the granularity of the time periods during preprocessing, thus eliminating the need for training time. We validate this approach with four virtual machines on the CDX 3.0 platform, each simulating different aspects of DDoS attacks for offensive, monitoring, and defense evaluation purposes. The study further analyzes seven diverse DDoS attack methods. When compared with existing methods, our approach demonstrates superior performance, highlighting its potential effectiveness in real-world DDoS attack detection.
分布式拒绝服务 (DDoS) 攻击对互联网和云安全构成重大威胁。我们的研究利用泊松分布模型，以 O() 的计算复杂度有效地检测 DDoS 攻击。与基于机器学习 (ML) 的算法不同，我们的方法仅在预处理期间根据时间段的粒度为合法流量设置一个或多个泊松模型，从而无需训练时间。我们使用 CDX 3.0 平台上的四台虚拟机验证了这种方法，每台虚拟机模拟不同方面的 DDoS 攻击，用于攻击、监控和防御评估目的。该研究进一步分析了七种不同的 DDoS 攻击方法。与现有方法相比，我们的方法表现出优越的性能，突出了其在现实世界 DDoS 攻击检测中的潜在有效性。
