The intrusion detection systems (IDSs) developed based on classification algorithms for securing wireless sensor networks (WSNs) are unable to attain the required detection accuracy. To handle the security issue in WSN, an intelligent IDS is proposed in this work by using a convolution neural network (CNN)-based deep learning approach along with a fuzzy inference model. The proposed IDS keeps track of the network and system activities by using the proposed fuzzy CNN along with spatial and temporal constraints to detect malicious nodes. Moreover, this algorithm has been modelled mathematically by using Feynman Path Integral and Schrodinger equation for handling the spatial and temporal constraints with fuzzy rules. From the experiments conducted in this work, it is proved that the proposed IDS increases the security, detection accuracy and packet delivery ratio, but decreases the delay and false positive rate in WSNs when compared with the existing IDSs.