基于BP神经网络的DDoS攻击检测研究

Research on DDoS detection based on BP neural network

摘要:

分布式拒绝服务攻击(DDoS)是如今常见的网络威胁之一，DDoS攻击易被发动却很难追踪与防范。在神经网络快速算法基础上，首先系统分析国内外DDoS攻击检测理论、方法与大量数据集，构建了基于数据包长度，数据包发送时间间隔以及数据包长度变化率等七项特征的攻击流量特征模型，其次通过大量尝试提出对神经网络误差调整参数进行优化的方法，最后基于加州大学洛杉矶分校数据集（UCLA CSD Packet Traces）进行了参数改进前后的攻击检测对比实验。实验表明，提出的方法能有效提高DDoS攻击检测率,且具有较好的泛化能力。

Abstract:

Distributed denial of service attack (DDoS) is a common threat in today’s networks. While such an attack is not difficult to launch, defending a network resource against it is disproportionately difficult. This paper analysis some famous theories and methods on detection of DDoS network attacks systematically based on the fast neural network algorithm. Meanwhile, the attack traffic feature model which is constructed based on the packet length, packet transmission time interval and packet length change rate etc is proposed. Second, a method to optimize the parameters of the neural network error is also proposed by a large number of attempts. Finally, the UCLA dataset is used to carry out the contrast experiment of the parameters before and after the improvement. Experiments show that the proposed method can effectively detect DDoS attacks and has a better generalization ability.