Integrated Deep Neural Networking Approach with Long Short-Term Memory (LSTM) for Bottleneck Detection in IoT Devices

Abstract

This review examines how machine learning (ML) is being used to spot bottlenecks in systems such as cloud computing, fog environments, and the Internet of Things (IoT). It covers some of the more recent developments, especially one standout technique: an amalgam deep learning model that combines DNN and LSTM algorithms to detect IoT botnet attacks. What’s striking is that this method manages to correctly identify 99.98% of even the most complex attacks, and it does so in just 0.022 milliseconds—a pretty solid performance. The paper also highlights the role of convolutional autoencoders, which have shown promise in detecting suspicious activity across IoT networks, hitting an impressive accuracy rate of 99.88%. Beyond that, it dives into some innovative approaches such as software-defined networking (SDN) and other ML techniques that are specifically geared toward managing and reducing botnet-related risks. There’s also discussion around newer feature selection methods and hybrid deep learning strategies that aim to boost memory efficiency while avoiding problems such as underfitting or overfitting. Overall, the review brings together a wide range of insights from recent research and points out practical tools and frameworks that could help tackle current and future cybersecurity issues across cloud, fog, and IoT systems. This approach is applied to 11 different IoT attacks with 20%, 50%, and 100% of the selected features. In this approach, an accuracy of 99.76% is achieved with 50% of the selected features, while an accuracy of 99.86% is obtained with 20% of the selected features.

Received: 1 July 2025 | Revised: 13 November 2025 | Accepted: 10 December 2025

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

The data that support the findings of this study are openly available in [“N-BaIoT”] at https://dx.doi.org/10.21227/y9de-qj71.