The internet of vehicles (IoV) is a network application for the internet of things (IoT) that allows smart automobiles to communicate with one other and with the public internet. Since the inception of IoV technology, there has been a significant increase in customer interest in smart automobiles. However, the rapid expansion of IoV has given rise to various security and privacy concerns, potentially leading to serious accidents. Numerous researchers have proposed intrusion detection models for IoT networks based on deep learning (DL) to mitigate smart car accidents and identify malicious attacks in vehicular networks. However, there is a demand for an effective and appropriate real-time algorithm for precisely recognizing malicious attacks in the IoV setting. In this study, a bidirectional long short-term memory (B_LSTM) model has been proposed and applied to the CAN_HCRL_OTIDS_UB dataset, sourced from the integrated vehicular network. Each neuron within the neural network retains weights to assign significance to information contributing to enhanced accuracy. To minimize loss functions and improve precision, the model incorporates various optimizers such as Adam, Adadelta, Adagrad, Adamax, AdamW, Adafactor, Nadam, and RMSprop. The results indicate that, when optimized with Adam, the B_LSTM model achieves superior prediction accuracy compared to alternative optimizers. Specifically, the suggested B_LSTM with Adam model demonstrates a high prediction precision of 0.9990 making it well-suited for effectively detecting attacks in the IoV environment.