基于深度强化学习的资源受限条件下的DIDS任务调度优化方法

在节点性能有限的边缘计算环境下进行分布式入侵检测系统(distributed intrusion detection system,DIDS)的任务分配,是一种典型的资源受限任务调度问题.针对该问题,提出基于深度强化学习的DIDS低负载任务调度方案.该方案将任务调度过程描述为马尔科夫决策过程(Markov decision process,MDP)并建立模型的相关空间和价值函数,找到保持DIDS低负载状态的最优策略.针对状态和动作空间过大且高维连续的问题,提出通过深度循环神经网络进行函数拟合.实验表明,所提出方案可使DIDS在网络变化中动态调节调度策略,保持系统整体的低负载,而安全指标没有明显降低.

An optimization method for DIDS task scheduling under resource- constrained conditions based on deep reinforcement learning

The task assignment of distributed intrusion detection systems(DIDS) in the edge computing environment with limited node performance is a typical resource-constrained task scheduling problem. To solve this problem, a DIDS low-load task scheduling scheme based on deep reinforcement learning is proposed. The task scheduling process is first described as a Markov decision process and the relevant space and value function of the model are established to find the optimal strategy for maintaining the low-load state of the DIDS. To solve the problem of excessively large action space and high-dimensional continuity, a deep recurrent neural network is proposed to perform function fitting. The experimental results show that the proposed scheme enables the DIDS to dynamically adjust the scheduling strategy during network changes, keeping the overall system load low, and the safety indicators are not significantly reduced.