Recently, Dr. Rufan Bai from our institute has published a paper entitled “On the Computation of Mixed Strategies for Security Games with General Defending Requirements” in the internationally renowned artificial intelligence journal Artificial Intelligence (AIJ, CCF-A). This work presents the team’s latest research advances in the field of multi-agent security games. Dr. Rufan Bai is the sole first author of the paper. This also marks the second time that our university has published in this journal as the first or corresponding affiliation.
Within the domain of multi-agent games, this study investigates how defenders allocate limited resources across network nodes to mitigate potential losses under both adversarial and uniform attack models. It highlights the advantages of mixed strategies over pure strategies and quantitatively evaluates their defensive performance under different attack scenarios. Notably, the work is the first to establish an approximate equivalence between optimal mixed strategies and optimal fractional strategies under a general threshold model. From a theoretical perspective, the study provides new insights into security game analysis, where defenders must make optimal decisions under strict resource constraints. While prior work has largely focused on pure strategies, this paper offers a systematic analysis of mixed strategies in a general framework. The results show that although mixed strategies can significantly improve defense performance in certain settings, computing the optimal solution is highly complex (NP-hard). To address this challenge, the authors propose a novel Patching algorithm, which efficiently computes mixed strategies with theoretical guarantees under adversarial attacks and demonstrates robustness to problem scale, offering a practical solution for real-world applications.
Founded in 1970, Artificial Intelligence is widely recognized as one of the most prestigious journals in the field of classical artificial intelligence and is classified as a CCF-A journal by the China Computer Federation. The journal publishes only around one hundred papers annually, with contributions from mainland China accounting for less than 10%. In recent years, supported by the National Natural Science Foundation of China (Young Scientists Fund) and the Jiangsu Provincial Natural Science Foundation (Young Scientists Fund), Dr. Rufan Bai has been actively conducting research in multi-agent games, fair allocation, and autonomous driving.
