An intelligent distributed network system is the Wireless Sensor Network (WSN), which is a strategy required to address security threats as well as energy consumption that has a direct impact on a network’s lifetime. Thus, attempting to identify malicious attacks with a low consumption of data transmission makes a lot of sense. The high energy consumption of nodes due to the transmission of data shortens the lifetime of the network. To overcome these issues, the proposed method is based on the Ordering Points to Identify Cluster Structure (OPTICS) with Bi-directional Long Short Term Memory using Differential evolution (DE-BiLSTM) classifier to detect the threats in WSN for smart building. Initial deployment of the sensor nodes (SN) and formation of the cluster nodes (CN) by employing the OPTICS density-based clustering approach that partitions clusters with different densities. In order to transport data to the base station, the cluster head (CH) nodes are chosen from the CN according to their more energy as well as shorter distance. Then, in order to forecast the threats, the size of the batch and hidden layers are set using the differential evolution method (DE) and the classification of the data is performed using BiLSTM to detect as attack or non-attack. Performance for predicting an attack is measured by network and classification parameters such as Packet Delivery Ratio (PDR), Average Residual Energy (ARE), Throughput, Accuracy and Precision. The results of the performance obtained are 91.78% for PDR, 8.56J for ARE, 2.52mbps for throughput with 100 nodes, then 93.78% for accuracy and 93.04% for precision. Thus, the designed detection of threats in WSN based on OPTICS clustering with DE-BILSTM classifier performs better for malicious attack prediction with low energy consumption sensor nodes.