A reduced analytical review of intelligent IoT security: Trends, gaps, and future challenges

The Internet of Things (IoT) connects small and low-power devices used in homes, hospitals, industries, and cities. These devices often face attacks such as botnets, denial-of-service, malware, intrusions, anomalies, and privacy leaks because of their limited memory and energy. This paper reviews recent studies and compares four types of data used for defense: network traffic, device activity logs, memory traces, and image-based features. Reported results show that many machine learning and deep learning methods achieve above 99% accuracy on public datasets, but their performance falls in real settings with encrypted traffic and non-uniform data. Promising directions include using compressed models that fit device limits, shifting heavy analysis to the network edge, applying federated learning with robust aggregation, and keeping blockchain at gateways while devices use only light clients. To guide fair evaluation, we recommend reporting memory use, energy, latency, and model size in all future studies.