Under a novel adaptive event-triggered communication scheme, a dual security control problem is investigated by combining mechanistic analysis with a data-driven approach for industrial cyber-physical systems(ICPS) with denial-of-service(DoS) attacks and actuator fault. First, in order to save more network communication resources, the trigger parameters of the designed adaptive event triggered scheme can change dynamically and adaptively with the system behavior. Second, an attack detection mechanism is established based on the maximum allowable delay of the system to effectively distinguish large and low energy DoS attacks. Third, based on extreme learning machine algorithm, a timing prediction model is established to reconfigure and correct the control quantity during high-energy DoS attacks, thus actively tolerating the impact of attacks, which is combined with resilient passive attack tolerance for low-energy attacks to improve the system's defense against attacks. Fourth, with the help of T-S fuzzy theory, time delay theory, and the new Bessel-Legendre inequality, etc., a robust observer and an analytical solution method of the dual security controller are obtained to improve and balance the performance of dual security control and communication. Finally, the effectiveness of the proposed method is verified by simulation examples.