<p>Security vulnerabilities such as intrusion attempts, zero-day attacks and anomalies have become primary issues as rapidly growing BIoT networks increase exponentially. Despite enormous false-positive ratios and computational overheads, legacy IDS like signature-based and anomaly-based systems cannot provide real-time accurate and adaptive detection. To enhance intrusion detection precision and scalability in BIoT environments, this research proposed a Hybrid Transformer-LSTM Deep Learning Framework that integrates the temporal analysis capabilities of LSTM with the feature extraction abilities of the Transformer model. Accuracy, precision, recall and F1-score are some of the significant performance metrics utilized to evaluate the proposed model after it has been trained on large network traffic datasets. The Hybrid Transformer-LSTM model outperforms prior methods such as Bi-LSTM, CNN and traditional ML-based IDS approaches with a detection rate of 98.53 percent according to experimental findings. The threat detection is made reliable by the low FPR = 0.0134 and FNR = 0.0159 rates of the model. Furthermore, the blockchain incorporated as a measure to enhance security and accountability ensures immutable logging of incursions detected. The scalability of blockchain and the encryption-decryption speed is also discussed, and it is possible to emphasize the possibility to deploy the suggested framework to large IoT networks. The experimental results confirm the fact that the application of a hybrid transformer-LSTM solution in the IDS significantly enhances security, accuracy and performance in the IoT applications supported by blockchain.</p>