Internet of Things (IoT) is the network connected among physical devices with various sensors which plays a key role in modern and smart societies. However, the IoT paradigm is prone to security concerns as many attackers try to hit the network and make it vulnerable. To overcome these issues, a novel Firefly optimization Algorithm combined SNN with Conv-BiLSTM for attack termination in IoT (FAST-IoT) approach is proposed for d etecting attacks in IoT environment. Initially, the IoT devices generate data packets, which undergo data pre-processing. The processed data is then passed through a Spiking Neural Network model (SNN) for feature extraction. These features are selected by using the Firefly Optimization Algorithm (FOA) which is fed into a Convolutional-Bidirectional Long Short-Term Memory (Conv-BiLSTM) model. The Conv-BiLSTM model classifies the input as either Attack or No-Attack and if an attack is detected, it is blocked otherwise, the user proceeds without interruption. The FAST-IoT approach is evaluated by using the N-BaIoT dataset and it is simulated by a cloud simulator (Cloudsim). The proposed FAST-IoT approach achieves the highest accuracy of 93.2%, compared to 65.8% for DeepAK-IoT, 78.6% for MalBoT-DRL, and 87.4% for Hierarchical Cloud DNN respectively.