The routing performance in Wireless Multimedia Sensor Network (WMSN), which transfers and receives multimedia content like a scalar, audio and video, is often affected by malicious nodes and residual nodes. An external attacker modifies the characteristics function of the node, and thus the node becomes malicious nodes in WMSN. These malicious nodes will affect the functionalities of its surrounding nodes and prevent routing through it and other nodes. Hence, the detection and mitigation of malicious nodes are essential to improve the routing efficiency in WMSN. The conventional methods mainly used machine learning algorithms to identify the malicious nodes in WMSN, which provided low accuracy and consumed more detection time as the main drawbacks. The proposed methodology resolves these drawbacks of the conventional algorithms in this paper. This paper presents an efficient method for detecting and mitigating the malicious nodes using feature index, which is optimized by a Genetic Algorithm (GA). The optimized feature set is classified by the modified LeNET deep learning classification approach. Even though conventional deep learning architectures provide a high classification rate for malicious node detection, it consumes a high detection time to identify malicious nodes. This drawback is overcome by modifying the internal layers of the existing LeNET architecture into parallel, and the dense layers in the existing LeNET architecture are replaced by Fuzzy C Means (FCM) algorithm. The performance of the proposed methodology is analyzed with respect to misclassification rate, precision, recall, accuracy and F1-score parameters.