The hidden failures generally exist in power systems and could give rise to cascading failures. Identification of hidden failures is challenging due to very low occurrence probabilities. This paper proposes a state-failure-network (SF-network) method to overcome the difficulty. The SF-network is formed by searching the failures and states guided by risk estimation indices, in which only the failures and states contributing to the blackout risks are searched and duplicated searches are avoided. Therefore, sufficient hidden failures can be obtained with acceptable computations. Based on the state and failure value calculations in the SF-network, the hidden failure critical component indices can be obtained to quantify the criticalities of the lines. The proposed SF-network method is superior to common sampling based methods in risk estimation accuracy. Besides, the state and failure value calculations in the SF-network used to re-estimate the risks after deployment of measures against hidden failures need shorter time in comparison with other risk re-estimation methods. The IEEE 14-bus and 118-bus systems are used to validate the method.