Temporal Pattern-Aware Intrusion Detection in Smart Homes via SCOA-Driven Feature Selection and Deep Learning

Strong intrusion detection schemes are required because smart homes' propagation of internet of things (IoT) devices has made them more vulnerable to cyber threats. Using a combination of intelligent data preprocessing, novel feature selection via single candidate optimization algorithm (SCOA), and classification through a recurrent neural network (RNN), this study presents an advanced intrusion detection in smart home (IDSH) framework. Included in IDSH dataset's thorough preprocessing pipeline are operations such as handling categorical data via encoding, removing irrelevant columns, normalizing numerical features, and strategically imputation of missing values. dataset comprises time-series activity and network traffic logs from smart home environments. To extract statistical and temporal properties, feature engineering techniques are also used. By optimizing a fitness function based on RNN classification accuracy and feature sparsity, SCOA is used to identify the most relevant subset of features, which enhances model efficiency and reduces computational complexity. In order to effectively capture the temporal patterns inherent in data on smart home behavior, the features that were chosen are fed into a recurrent neural network (RNN). With an accuracy of 96.85% in binary intrusion detection and 92.56% in multiclass classification scenarios, the proposed model achieves high classification performance. Additional evaluation metrics that confirm the framework's robustness include recall, f1-score, area under the curve (AUC-ROC), and average precision. When compared to more conventional classifiers and feature selection techniques, the SCOA-RNN pipeline clearly comes out on top. A smart, scalable solution for protecting smart home ecosystems is presented in this work, which opens the door to low-latency, real-time threat mitigation.