Journal of Ad-hoc Network and Mobile Computing (e-ISSN: 3048-9180)

Adaptive QoS-Aware Routing in Mobile Ad-Hoc Networks Using Hybrid Machine Learning Techniques

Ismail Olaniyi Muraina, Bashir Oyeniran Ayinde — Sat, 17 Jan 2026 00:00:00 +0000

The Mobile Ad-Hoc Network (MANETs) use is vitiated by an ongoing issue of providing stable, effective and viable routing due to the extremely dynamic topology, the spatial distribution of nodes, the frequentness of link failures, and the constraints of energy and bandwidth. The above challenges make it problematic in a way that the capability of the traditional routing protocols to fulfill Quality of Service (QoS) requests on a consistent basis is difficult to achieve particularly in a high mobility environment with dense node distributions. To address these deficiencies, this paper proposes a hybrid machine learning-based routing framework that dynamically selects the most appropriate routing path as well as offers QoS-aware network performance. The proposed framework integrates supervised learning techniques to recreate the stability of the links and forecast the traffic patterns, and reinforcement learning techniques that can be used to generate proactive and intelligent routing decisions. Based on the information on past and current network states, the hybrid approach adapts itself to changes of topology, congestion prediction and dynamically selected routings that optimize different QoS parameters are selected. This collaborative learning system allows the network to compute routing decisions in a self-adaptable and intelligent manner, and remove any reliance on fixed or reactive routing protocols. The usefulness of the suggested methodology is experimented with the aid of a great number of simulations with the NS-3 network simulator, a set of diverse mobility models, node densities, and traffic scenarios. The key performance indicators, which are the end-to-end delay, ratio of delivered packets, throughput, and energy consumption, are acquired and compared with the traditional QoS-aware routing protocols. The hybrid learning based solution, as pointed out in the simulation results, has a major impact on increasing the degree of reliability of the network, latency, and energy efficiency, particularly in a highly mobile and densely populated MANET. This paper, by and large, proposes a scalable and sound routing structure, which bridges the gap between intelligent decision-making, and practical support of QoS of next-generation MANETs.

A Comprehensive Review of Graceful Degradation in Remote Healthcare Adhoc Wireless Sensor Networks

Manas Kumar Yogi, Nadiminti Sai Priya Satwika — Sat, 07 Feb 2026 00:00:00 +0000

Remote healthcare monitoring systems rely heavily on Adhoc Wireless Sensor Networks to provide continuous, real-time data on patient health, enabling proactive medical interventions and chronic disease management. However, these networks are often deployed in unpredictable, dynamic environments and are subject to a wide range of failures, including sensor malfunctions, battery depletion, and network congestion. Graceful degradation the ability of a system to maintain its most essential functions even when some components fail is critical for ensuring patient safety and data integrity in these high-stakes medical applications. This review article provides a comprehensive and investigative analysis of graceful degradation strategies in remote healthcare WSNs, focusing on recent advancements and technological innovations from 2021 to 2026. The authors examine the fundamental architectural requirements for resilience, the multi-stage fault tolerance mechanisms for medical sensors, and the integration of emerging technologies such as Artificial Intelligence (AI), edge computing, and the Internet of Things (IoT) to enhance system robustness. By synthesizing findings from popular recent studies, this article identifies key challenges, such as the energy-reliability trade-off and security vulnerabilities, and outlines future directions for developing more dependable, adaptive, and human-centric remote monitoring systems. The findings underscore the critical importance of multi-modal sensing, decentralized management, and explainable degradation in achieving high reliability and seamless performance under adverse conditions, ultimately ensuring that the care for the patient remains uninterrupted regardless of technical challenges.

Deep Learning-Based Context-Aware Resource Allocation for Networked Mobile Computing Environments

N. B. Mahesh Kumar — Mon, 30 Mar 2026 00:00:00 +0000

This paper proposes a deep learning-based, context-aware resource allocation framework for networked mobile computing environments, addressing dynamic challenges like fluctuating networks, user mobility, device heterogeneity, and energy constraints. The approach utilizes a multidimensional context vector encompassing device status (CPU, memory, battery), network conditions (bandwidth, latency), user mobility patterns, and application requirements (task size, complexity, deadlines) to dynamically decide between local task execution and offloading to edge/cloud servers. A fully-connected deep neural network (DNN) with 3-4 hidden layers (128-16 neurons, ReLU activations) and sigmoid output approximates the multi-objective optimization problem of minimizing latency and energy consumption, trained via supervised learning on 10,000 simulated scenarios using binary cross-entropy loss. NS-3 simulations demonstrate superior performance: 4-11% latency reduction over DRL/LSTM baselines, 43% energy savings versus context-unaware methods, and up to 50% higher throughput across 50-200 devices, with ablation studies confirming network context's criticality. The framework's novelty lies in its comprehensive context integration, low-overhead inference (5 ms), scalability, and detailed architecture, outperforming rule-based, DRL, and LSTM approaches for real-world mobile applications.

Design and Implementation of an AI-Based 3D Crime Scene Reconstruction from Visual Evidence

Pragna C. P, Keerthana B, Lekhana K, Sunayana Anuja, Priyanka H. V — Mon, 13 Apr 2026 00:00:00 +0000

Artificial Intelligence (AI) and computer vision have significantly improved the performance of object detection systems. However, most existing approaches generate only two-dimensional (2D) outputs, limiting the understanding of spatial relationships between objects. This paper presents a system that integrates deep learning-based object detection with three-dimensional (3D) visualization to address this limitation. The proposed method utilizes the YOLOv5 model for accurate object detection and the Open3D library to construct an interactive 3D environment. Detected objects are mapped onto textured planes and arranged within a 3D space, enabling users to explore the scene from different perspectives. This enhances the interpretation of visual data compared to traditional 2D representations. The system is designed to be simple and efficient, requiring only image input to generate a 3D visualization. Experimental results demonstrate improved understanding and presentation of object detection outcomes. Overall, the proposed approach provides a more intuitive and effective solution for analysing visual data, with potential applications in surveillance, research, and smart monitoring systems.

A Practical Approach to Man-in-the-Middle (MitM) Attacks Detection in IoT Networks

Abhishek D., Durgesh Nishad, Nayan Gowda D., Ganavi M. R., Abirami A. — Thu, 16 Apr 2026 00:00:00 +0000

The rapid development of the internet of things (IoT) has significantly transformed the contemporary communication system, which connects a vast array of intelligent devices. This growth has, however, also posed severe security lapses occasioned by low computational abilities and low security measures in the IoT devices. The Man-in-the-Middle (MitM) attack is one of the most severe attacks, in which a vulnerable person secretly intercepts and modifies communication between two devices without their awareness. This research work is aimed at the analysis and visualisation of the MitM attacks within a controlled IoT network setting. A real-life practice is followed using tools like Kali Linux, Wireshark and ARP spoofing techniques to help in simulating a real-life attack scenario. The experiment demonstrates the way attackers may intercept sensitive information, including login credentials and network data, by putting themselves between communicating devices. The findings demonstrate the extreme dangers of unsecured communication schemes in the IoT systems. Moreover, the study will discuss effective detection and prevention measures, such as encryption, secure communication protocols, intrusion detection systems, and network monitoring measures. This study has highlighted the need to enhance IoT security systems in order to ensure data integrity, confidentiality, and reliability of the system. It is important to deploy strong security tools that will protect the IoT environment against the emerging cyber threats, such as MitM attacks.