Control of a discrete-time linear system over an insecure wireless network is investigated. Specifically, the channel used for transmission of control input packets is assumed to be subject to jamming interference attacks. In this setting, the likelihood of transmission failure at each time depends on the power of the interference signal emitted by the attacker at that time. We extend our previous work to consider the scenarios where the attacker uses the information of the system state and the dynamics to efficiently change the interference power and cause instability. We analyze the stability of the networked control system by investigating the properties of a martingale that depends on the transmission failure indicator and the interference power process. We obtain sufficient stability conditions, which indicate that closed-loop stability can be guaranteed if the attacker has energy constraints so that the average jamming interference power has a sufficiently small upper bound. The effect of state-dependent jamming and the utility of our analysis approach are illustrated through an attack strategy with rolling-horizon optimization, where the attacker decides the interference power based on maximizing a utility function that involves the predicted future states given the present state information.