Cubic-DUCAG: A New Modeling and Probabilistic Computing Approach for Cyclic Network Attacks

By calculating the risk probabilities of nodes within the attack graph, the attack paths and scenarios can be constructed to effectively describe the attack process. The general the probabilistic computing approaches for attack graph analysis require that the graph does not contain cycles, for that the cycles will bring about repetitive calculations and increased complexity, resulting in inaccurate results. However, loop attack paths often occur in the network obtained by applying the attack graph generator such as MulVAL. To solve the above issues, this study presents the Cubic-Dynamic Uncertain Causality Attack Graph (Cubic-DUCAG) for modeling the cyclic network attacks, and an algorithm for calculating the risk probability of node in the cycle paths based on causal chains (Cycle-CCRP). The proposed method offers a unified framework for calculating the risk probability of nodes within cycles while without changing the attack graphs. Experiments show that the Cycle-CCRP algorithm exhibits superior time efficiency characteristics compared to the Variable Elimination algorithm in Bayesian attack graphs and demonstrates good scalability in large-scale computer networks.