پریسا رحمانی

Background: Network security and penetration pose a significant challenge in the extensive IoT research of recent years. System security and user privacy demand security solutions that are carefully planned and diligently maintained. Aims: This paper introduces a novel three-stage hybrid IDS, IoT-APOSCA, leveraging machine learning and meta-heuristics for attack detection; stages include pre-processing, feature selection, and attack detection. The pre-processing steps are: cleaning, visualization, feature engineering, and vectorization. Methodology: Networks use Intrusion Detection Systems (IDSs) to monitor and detect malicious activities as a key security feature. The Arctic Puffin Optimization (APO) and Sine-Cosine Algorithm (SCA) are used in the feature selection stage, while a changed Random Neural Network (RNN) is employed in the attack detection stage. Results: The proposed technique is assessed using the DS2OS dataset, and the outcomes show that the approach, integrating multiple learning models, led to an accuracy enhancement to 99.66%. Also, the values Recall and False Alarm Rate obtained are equal to 0.9926 and 0.003, respectively. Conclusion: Intrusion detection system efficacy is directly tied to the quality of its classification method. Enhanced neural network performance is achievable through adjustments to parameters, such as network weights.