SAFNet-IoT: Secure Adaptive and Cloud supported Federated Network for IoT-Based Industrial Anomaly Detection

It is a great need to enhance the robust, secure and efficient anomaly detection mechanisms in the context of fast expanding Industrial Internet of Things (IIoT) systems. Over the years, numbers of traditional centralized models have faced lots of challenges such as risks to data privacy, high communication overhead, and vulnerability to adversarial attacks. However, 1 there are limitations which we address with the new federated learning framework termed as SAFNet-IoT (Secure Adaptive Federated Network for IoT) that encompasses Blockchain Based Authentication (BBA) and an Adaptive Autoencoder LSTM (AAEL) anomaly detection module. Smart contracts are used in the BBA mechanism to perform the validation and security of the model updates, by preventing adversarial modifications and maintaining the data integrity. At the same time, the AAEL module dynamically adapts to hyperparameters in response to real time feedback, and at a given instant in time optimize the anomaly detection within heterogeneous IIoT environments. We showcase a SER however, that SAFNet IoT achieves 94.7% anomaly detection accuracy, better than these other conventional FL models, i.e., FedAVG-LSTM (93.1%) and Deep Autoencoder (91.2%). Another thing is that the framework reduces communication overhead by 30% with a bandwidth reduction ratio (BRR) of 0.67, which results in greater scal-ability. Moreover, SAFNet-IoT enables 1.4 seconds average local training time and 4.6 seconds system latency, which is computationally efficient for resource-constrained IIoT nodes. By connecting to the blockchain, security is improved and the authentication success rate is 98.5% while 91.2% of malicious updates are detected outperforming traditional secure aggregation methods. This finding shows that SAFNet-IoT is effective in enhancing the anomaly detection, security resilience as well as communication efficiency in IIoT scenarios. Work on future optimization of smart contract execution, minimizing costs of blockchain transaction, and development of SAFNet-IoT into multi-modal data fusion IIoT anomaly detection will follow.