Learning‐driven ubiquitous mobile edge computing: Network management challenges for future generation Internet of Things
Praveen Kumar Donta, Edmundo Monteiro, Chinmaya Kumar Dehury, Ilir Murturi · International Journal of Network Management · 2023
Ubiquitous edge computing facilitates efficient cloud services near mobile devices, enabling mobile edge computing (MEC) to offer services more efficiently by presenting storage and processing capability within the proximity of mobile devices and in general IoT domains.However, compared with conventional mobile cloud computing, ubiquitous MEC introduces numerous complex challenges due to the heterogeneous smart devices, network infrastructures, and limited transmission bandwidth.Processing and managing such massive volumes of data generated from these devices is complex and challenging in edge infrastructures.On the other side, time-critical applications have stringent requirements such as ultra-low-latency, energy cost, mobility, resource, and security issues that cannot be neglected.For example, smart healthcare or industrial networks generate emergency information very frequently (i.e., often in terms of milliseconds), which needs to be processed near the sensing devices with minimal processing delay.In this context, future generation IoT requires robust and intelligent network management approaches that can handle the system complexity (e.g., scalability and orchestration) with little or no little human intervention and offer a better service to end-users.More precisely, AI/ML approaches designed explicitly for networks under high traffic volume of data help overcome several management challenges, such as (i) improving performance by balancing load and traffic, (ii) distributing the bandwidth spectrum based on demand, and (iii) traffic predictions.Moreover, this need also opens several new research directions such as new MEC architecture, service provisioning technique, security mechanisms, advanced 5G or beyond communication technology, ambient intelligence, and AI/ML-based solutions.This issue collect surveys and contribution articles on emerging trends and technologies in ubiquitous MEC for future generation IoT networks and their managements.The papers related to machine learning, deep learning, optimization, blockchain, 5G, or beyond solutions, especially for domain-specific IoT network management, which use MEC environments, are collected after evaluating the review process.Each paper submitted to this special issue was reviewed by three to seven experts during the assessment process.At the end we consider one survey paper and four research contributions.The first paper Ravi et al. proposed a survey on "Stochastic modeling and performance analysis in balancing load and traffic for vehicular ad hoc networks."This survey presents recently published stochastic modeling-based algorithms for VANETs.This article briefly covers various queueing models for the reader's convenience.This paper discusses a variety of VANET issues such as mobility, routing, data dissemination, cooperative communication, congestion control, and traffic load balancing issues addressed by stochastic modeling techniques.The authors provided extensive open challenges for young researchers on several network management topics such as zero-touch provisioning, blockchain, intelligent digital twin, cooperative communication among vehicles, and routing predictions according to vehicle mobility.These topics are emerging in this area, and they are open to further research.The second paper Song et al. contributed on "A group key exchange and secure data sharing based on privacy protection for federated learning in edge-cloud collaborative computing environment."The proposed work aims to secure the transmission of model parameters between IoT terminals during a federated learning process.With the key selfverification algorithm, the model legitimizes the public and private keys of the terminal, which greatly enhances their security.In order to protect against terminal identity leakage, Song et al. provide an attribute-based cryptographic method; the terminal generates ciphertext attributes and gives them to the cloud server.The security of sharing the parameters of each model in the FL process is guaranteed.Terminal self-adaptive is achieved by defining access structures to meet shared resource access rights.Upon satisfying the access authority, each terminal stores and downloads the shared ciphertext model parameters from the edge-cloud server.