Title: Security Defense of Large Scale Networks Under False Data Injection Attacks: An Attack Detection Scheduling Approach Show Chinese title

Title: 大规模网络在虚假数据注入攻击下的安全防御：一种攻击检测调度方法

Abstract: In large scale networks, communication links between nodes are easily injected with false data by adversaries, so this paper proposes a novel security defense strategy to ensure the security of the network from the perspective of attack detection scheduling. Compared with existing attack detection methods, the attack detection scheduling strategy in this paper only needs to detect half of the neighbor node information to ensure the security of the node local state estimation. We first formulate the problem of selecting the sensor to be detected as a combinatorial optimization problem, which is Nondeterminism Polynomial hard (NP-hard). To solve the above problem, we convert the objective function into a submodular function. Then, we propose an attack detection scheduling algorithm based on sequential submodular maximization, which incorporates expert problem to better cope with dynamic attack strategies. The proposed algorithm can run in polynomial time with a theoretical lower bound on the optimization rate. In addition, the proposed algorithm can guarantee the security of the whole network under two kinds of insecurity conditions from the perspective of the augmented estimation error. Finally, a numerical simulation of the industrial continuous stirred tank reactor verifies the effectiveness of the developed approach. Show Chinese abstract

Abstract: 在大规模网络中，节点之间的通信链路容易被对手注入虚假数据，因此本文提出了一种新的安全防御策略，从攻击检测调度的角度确保网络的安全性。 与现有的攻击检测方法相比，本文的攻击检测调度策略只需检测一半的邻居节点信息即可确保节点局部状态估计的安全性。 我们首先将选择被检测传感器的问题形式化为一个组合优化问题，该问题是非确定性多项式难（NP-hard）。 为了解决上述问题，我们将目标函数转换为子模函数。 然后，我们提出了一种基于顺序子模最大化攻击检测调度算法，该算法结合专家问题以更好地应对动态攻击策略。 所提出的算法可以在多项式时间内运行，并具有优化率的理论下限。 此外，从增强估计误差的角度出发，所提出的算法可以在两种不安全条件下保证整个网络的安全性。 最后，对工业连续搅拌釜反应器的数值仿真验证了所开发方法的有效性。