Risk-averse bi-level stochastic network interdiction model for cyber-security risk management

•The first study to model a risk-averse cyber network defender over an attack graph.•Develop a risk-averse bi-level stochastic network interdiction model.•Develop a customized exact algorithm with novel acceleration techniques.•Mean-risk stochastic model substantially outperforms the risk-neutral and deterministic models.•Optimal interdiction policy changes as the risk-preference varies.