Deep Guard-IoT: A Systematic Review of AI-Based Anomaly Detection Frameworks for Next-Generation IoT Security (2020-2024)

Abstract

The emergence of IoT devices has complicated the landscape of cybersecurity in ways that had never been experienced before, thereby, giving raise to the need for more developed methods that can assist in threat evaluation and deterrent. The work in the international scope analyses the particulars of the artificial intelligence-based anomaly detection in the IoT technology implementation including the perspectives of recent period between 2020 and 2024 of the various architectures’ effectiveness and their use in low-resource IoT environments. In this review, as well as in many other recent works carried out by other authors, it is observed that deep learning methods, such us Long Short-Term Memory (LSTM) networks and GRU-LSTM hybrid models, achieved the most accurate performance, ranging from 96% up to 99.9% correct detection. Our examination focuses on several aspects of IoT security, such as challenges at the device level, issues related to security at the network level, data security, and facets related to artificial intelligence concepts and architectures capable of addressing the challenges. The results of the study state that the AI techniques tend to be more efficient and have a high performance than the conventional methods before but have drawbacks for realistic application owing to lack of adequate resources, absence of standard practices and sophistication of new threats. The study also highlights the gaps that exist in addressing the current approaches and makes suggestions on what is to be done, such as the importance of developing