Akademik Veri Yönetim Sistemi
PEERJ COMPUTER SCIENCE, cilt.12, ss.1-27, 2026 (SCI-Expanded, Scopus)
The quick spread of Internet of Things (IoT) devices in smart homes has raised cybersecurity concerns, calling for smart, flexible, and active ways to reduce threats. Standard Intrusion Detection Systems (IDS) and basic anomaly detection have trouble spotting new attacks and changing cyber threats. To fix these problems, this article puts forward a Reinforcement Learning-Based Adaptive Threat Mitigation (RL-ATM) model. It uses methods like reinforcement learning, deep learning, and data mining to make smart home cybersecurity better by reducing risks and finding network issues. Tests show that RL-ATM does much better than current cybersecurity options, such as signature-based IDS, anomaly-based machine learning models, and deep reinforcement learning (DRL) setups. The model got an accuracy of 98.87%, a precision of 97.49%, a recall of 98.36%, and a low false positive rate (FPR) of 1.8%. This makes it a dependable cybersecurity choice for actual smart home use. A comparison shows that standard IDS models are only 87.42% accurate with an FPR of 6.3%. Anomaly-based ML methods improve accuracy to 91.15% but still have an FPR of 4.9%. Hybrid Convolutional Neural Network (CNN) + Reinforcement Learning models reach 92.84% accuracy but can’t adapt to changing attacks in real time. This makes RL-ATM a better choice for dependable detection and response. This work adds to smart home cybersecurity by giving a scalable, adaptive, and independent artificial intelligence (AI)-driven security system. It can lower cyber threats in real time.