It is crucial to improve the robustness of the IoT against malicious attacks. For scenarios where the network topology is split due to malicious attacks, but can still reconnect fragmented networks by adjusting links between nodes, existing algorithms offer limited robustness improvements, exhibit high computational complexity, incur significant energy consumption, and produce optimized network topologies lacking scalability. To address these issues, we propose a network robustness optimization algorithm called the ’Recovery Algorithm’. Based on this algorithm, the IoT can automatically adjust its topology to effectively resist malicious attacks. Compared with other algorithms, the optimized network of this algorithm possesses higher robustness and higher communication efficiency without affecting the scalability of the original network. Furthermore, the computational complexity and energy consumption of this algorithm are several orders of magnitude lower than existing algorithms.

Cheeger’s inequality is used to estimate a lower bound on the number of edges that need to be cut off to split a regular graph into two subgraphs with the same number of nodes. Therefore, we design a network topology that is robust under malicious attacks based on Cheeger’s inequality. Considering that Cheeger’s inequality cannot be directly used to optimize complex networks in reality, we generalize the conclusion of Cheeger’s inequality and use this conclusion to propose the Good Regular Graph algorithm (GRG) for generating strongly robust network topologies under malicious attacks. In addition, we explain the design principle of the algorithm from a graph-theoretic perspective and give the corresponding network topology expansion algorithm. GRG, after appropriate processing, can be applied to optimize the substructures of a wide range of real-world network topologies to enhance their robustness. Experiments show that the algorithms proposed in this research perform better compared to other algorithms. We prove from both theoretical and experimental perspectives that Cheeger’s inequality can serve as a basis for designing robust network topologies under malicious attacks.