ADAPTIVE CONFIDENCE-BASED DECISION MAKING IN DEEP LEARNING FOR CYBER THREAT DETECTION AND MITIGATION IN IoT

DOI: https://doie.org/10.10399/JBSE.2025517916

Thanuja Narasimhamurthy, Gunavathi Hosahalli Swamy

Keywords:

MIMO Internet-of-Things, Artificial Intelligence, Machine Learning, Deep Learning, Cyber threats.

Abstract:

Cyber threat has been always a critical concern in Internet-of-Things (IoT) while the problem is still an open-ended. Review of existing literatures shows significant contribution of Artificial Intelligence (AI) in form of Machine Learning (ML) and Deep Learning (DL) approaches. Yet the shortcoming in security has not been mitigated. In the line of addressing this problem, this paper introduces an AI-driven intrusion detection system capable of identifying the nature of inbound traffic in IoT. The system uses Random Forest for selecting most potential feature while Convolution Neural Network is used on ranked feature for further optimization in prediction process based on empirically evaluated confidence score. Assessment with benchmarked UNSW-NB15 dataset shows proposed system with 94.25% accuracy, 91.96% of F1-Score, and 1.93s of response time which are significantly better in contrast to frequently adopted AI models in literatures to exhibit highly secure and computationally cost-effective solution in IoT.

PDF