A Min-Max Normalized Self-Attention Network with Scale-Aware MEET Transformer for IoT Attack Detection

Data collection and sharing in the healthcare from the Internet of Things (IoT) makes it easy. But, IoT systems still can be attacked so strong attack detection systems are needed. The Scale-Aware Meet Transformer with Min-Max Normalization Self Attention Network (SAMT-MMN-SANet) is a sophisticated method for identification of cyber security threats in the context of IoT. To ensure efficient data normalization, the methodology starts with pre-processing the raw UNSW-NB15 dataset using Min-Max Normalization (M-MN). The Scale-aware Modulation Meet Transformer (SMMT) then uses an arbitrary-size sliding window to extract computational features from the normalized data. An attack detection model using the Spatial Frequency Self Attention Network (SF SANet) is enhanced with an error lowering method and the accuracy is improved. Incentives are provided in a configurable way using an event repository and a dynamic reward function, and action results from this approach can be conditioned on for iteratively training. The Gold Rush Optimizer (GRO) which implements these principles improves system robustness and consequently system efficiency. Experimental results demonstrate that the SAMT-MMN-SANet based detection of attacks system detects attacking with an extraordinary accuracy of 99.9% while significantly reducing detection time, computation overhead and error rate compared to traditional reinforcement learning methods. The method presented herein for reducing cyber threats in Internet of Things applications provides a solid and strong basis.