Critical review of Botnet sophistication and detection inefficiencies in IoT and mobile networks

Botnets are continuing to be a significant contributor to large-scale cyberattack, and their detection in IoT and mobile ecosystems is progressively challenging because of limited resources, encrypted messaging, command and control strategy development, and concept drift. The review will cover botnet detection techniques on IoT and mobile networks using machine learning, deep learning, SDN-enabled, hybrid, and explainable-AI techniques. Instead of this, we take into account the deployment factors, such as the latency, memory footprint, scalability, and robustness to evasion and obfuscation, which are supported by the evidence presented in the literature. Our recurrent constraints are summarized; among them are the use of legacy or non-representative data, costly feature pipelines, lack of real-time scalability, and lack of interpretability to be used operationally. On the findings of these, we define research directions to adaptive, drift-resilient, privacy-preserving, and explainable pipelines to detection which can be applied to heterogeneous, resource-constrained environments. It is important to note that all the performance numbers mentioned are reported in the original research and cannot be considered the outcome of one common testbed.