Blockchain and machine learning assure data, resilience, and intelligent threat detection, including blockchain and machine learning. The tamper-resistant blockchain protects sensitive information with a decentralized storage mechanism; machine learning adds the’ smartness’ in real-time concerning emerging risks in this home. However, a more systematic framework that integrates secure storage mechanisms and streamlined, accurate classification of IoT data must be developed to enhance threat detection capabilities. To bridge this gap, the present work presents BLESS, a hybrid model designed explicitly for IoT security within a smart home combining Binary Particle Swarm Optimization (BPSO) and African Vulture Optimization Algorithm (AVOA) for feature selection, and as the fitness evaluator, it uses a Support Vector Machine. For secure data storage, BLESS has used a Blockchain Inter Planetary File System (IPFS) server so that management is decentralized and has tamper-proofing of IoT data. The model was tested using two prominent IoT datasets, NSL-KDD and UNSW-NB15. It shows an accuracy of 98.97% and 97.75%, which surpasses standalone BPSO and AVOA in accuracy, precision, recall, specificity, and F1-score. All these results show that BLESS is very effective for robust adaptive threat detection. Future work would be in the direction of scaling BLESS for other applications in IoT and improving feature selection with enhanced capabilities toward adaptability and accuracy.