A multi-objectives framework for secure blockchain in fog–cloud network of vehicle-to-infrastructure applications

Knowledge-Based Systems 290 (2024) 111576 Available online 29 February 2024 0950-7051/© 2024 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Contents lists available at ScienceDirect Knowledge-Based Systems journal homepage: www.elsevier.com/locate/knosys A multi-objectives framework for secure blockchain in fog–cloud network of vehicle-to-infrastructure applications Abdullah Lakhan a,j,k, Mazin Abed Mohammedb,j,k, Karrar Hameed Abdulkareemc,d, Muhammet Deveci e,f,g,∗, Haydar Abdulameer Marhoonh,i, Jan Nedoma j, Radek Martinek k,l aDepartment of Cybersecurity and Computer Science, Dawood University of Engineering and Technology, Karachi City 74800, Sindh, Pakistan bDepartment of Artificial Intelligence, College of Computer Science and Information Technology, University of Anbar, Anbar, 31001, Iraq cCollege of Agriculture, Al-Muthanna University, Samawah 66001, Iraq dCollege of Engineering, University of Warith Al-Anbiyaa, Karbala 56001, Iraq eDepartment of Industrial Engineering, Turkish Naval Academy, National Defence University, 34942 Tuzla, Istanbul, Turkey fThe Bartlett School of Sustainable Construction, University College London, Gower St, London, WC1E 6BT, UK gDepartment of Electrical and Computer Engineering, Lebanese American University, Byblos, Lebanon hCollege of Computer Sciences and Information Technology, University of Kerbala, Karbala, Iraq iInformation and Communication Technology Research Group, Scientific Research Center, Al-Ayen University, Thi-Qar, Iraq jDepartment of Telecommunications, VSB-Technical University of Ostrava, 70800 Ostrava, Czech Republic kDepartment of Cybernetics and Biomedical Engineering, VSB-Technical University of Ostrava, 70800 Ostrava, Czech Republic lFaculty of Electrical Engineering, Automatic Control and Informatics, Opole University of Technology, 45-758, Opole, Poland A R T I C L E I N F O Dataset link: https://github.com/ABDULLAH-R AZA/Blockchain-Socket,https://github.com/A BDULLAH-RAZA/Assignment-/blob/master/Bl ockchain-Malware-Detection.csv Keywords: Security Concave and convex Cyber-attacks Blockchain Cloud LSTM V2I Vehicular NSGA-II A B S T R A C T The Intelligent Transport System (ITS) is an emerging paradigm that offers numerous services at the infrastructure level for vehicle applications. Vehicle-to-infrastructure (V2I) is an advanced form of ITS where diverse vehicle services are deployed on the roadside unit. V2I consists of distributed computing nodes where transport applications are parallel processed. Many research challenges exist in the presented V2I paradigms regarding security, cyber-attacks, and application processing among heterogeneous nodes. These cyber-attacks, Sybil attacks, and their attempts cause a lack of security and degrade the V2I performance in the presented paradigms. This paper presents a new secure blockchain framework that handles cyber-attacks, as mentioned earlier. This paper formulates this complex problem as a combinatorial problem, encompassing concave and convex problems. The convex function minimizes the given constraints, such as time and security risk, and the concave function improves performance and accuracy. Therefore, numerous constraints, such as time, energy, malware detection accuracy, and application deadlines, require optimization for the considered problem. Combining the jointly non-dominated sorting genetic algorithm (NSGA-II) and long short-term memory (LSTM) schemes is the best way to meet the problem’s limitations. In this study, the paper designed a malware dataset with known and unknown malware. The different kinds of malware lists (e.g., cyber-attacks) are considered in the form of known and unknown malware lists with the characteristics, size of code, where malware comes from, attack on which data, and current status of the workload after being attacked by the malware. Our main idea is to present blockchain, NSGA-II, and LSTM schemes that handle phishing, routing, Sybil, and 51% of cyber-attacks without compromising application performance. Simulation results show that the study reduces delay and energy, improves accuracy, and minimizes security risks for vehicular applications. 1. Introduction The Intelligent Transport System (ITS) is an effective paradigm for public transport vehicular applications. Vehicular-to-infrastructure (V2I) is an efficient paradigm of ITS that offers cloud-based transport ∗Corresponding author at: Department of Industrial Engineering, Turkish Naval Academy, National Defence University, 34942 Tuzla, Istanbul, Turkey. E-mail addresses: [email protected] (A. Lakhan), [email protected] (M.A. Mohammed), [email protected] (K.H. Abdulkareem), [email protected] (M. Deveci), [email protected] (H.A. Marhoon), [email protected] (J. Nedoma), [email protected] (R. Martinek). services on the roadside unit side [1]. Transportation applications such as metros, taxis, buses, and trains are often called vehicular applications. Cloud computing services play a vital role in V2I and bring services to the network’s edge for vehicular applications. Fog computing is a subset of cloud computing that can be implemented https://doi.org/10.1016/j.knosys.2024.111576 Received 25 December 2023; Received in revised form 13 February 2024; Accepted 25 February 2024 Knowledge-Based Systems 290 (2024) 111576 2 A. Lakhan et al. at the roadside with minimum latency. The V2I paradigm allows applications to offload workloads to fog computing for processing with minimum latency [2]. V2I is a set of technologies that allow vehicles and roadside infrastructure to communicate with each other to perform their tasks [3]. In V2I, applications perform various tasks using unicast and broadcast communication models, such as road signals, location tracking, traffic lights, and accessing different fare and vehicle searching services with minimum congestion. V2I supports mobility and omnipresence, as well as ubiquitous services [4]. The V2I paradigm supports fully autonomous and semi-autonomous vehicles in the network [5]. However, in addition to the benefits of V2I for vehicle applications, there are many security standards and performance issues due to heterogeneous computing nodes and resources. Many studies have suggested secure V2I paradigms for vehicle applications. In these paradigms, applications use authentication-based schemes to offload and access secure data [1,2]. However, these studies only considered vehicle registration in V2I paradigms and extracted the anonymity of the vehicle during registration. These studies [3–5] suggested secure wireless connections and message authenticationenabled service schemes in V2I. For instance, traffic, light, and road availability for vehicular applications. However, these paradigms only focused on the application-level quality of services in terms of security and privacy in V2I. The secure networking enabled V2I presented [6] with security protocols and raised awareness about attackers with homogeneous nodes. The heterogeneous nodes with cloud computing and fog computing-based machine learning enabled secure scheduling for vehicular applications in V2I [7]. However, these studies only considered the specific node security and fraud detection behaviors for applications. The blockchain-enabled V2I paradigm presented [8] for heterogeneous nodes with the secure proof-of-work (PoW) method. The blockchain transactions between nodes are immutable and verified by algorithms in V2I. Blockchain is a new technology that offers immutable, transparent, and valid transactions for transport applications among heterogeneous nodes. A blockchain in the network is a collection of blocks based on the link-list data structure and data hashing based on cryptographic techniques. Each block has a digital signature, and the consensus algorithm validates node transactions. Blockchain is a distributed data structure called a distributed ledger (DL) [9]. It comprises a chain of linked blocks that hold all the time-stamped transactions on the network [10]. Every block in the tamper-proof ledger refers to the previous block’s hash. Different blockchain frameworks exist, such as private, public, and community. Each user and node in the network must be authenticated and checked before being part of the blockchain. A permission blockchain makes it less likely for users to share data that is not real, which can happen in public blockchain frameworks [11]. Blockchain played a vital role in terms of a single point of failure, centralized authenticity of the nodes for transport applications [12]. The public blockchain is permissionless, meaning that any anonymous user can be part of the blockchain, where every block is immutable, transparent, and tamper-proof. The community blockchain is like a private blockchain where all nodes can share their data and authenticate before adding it to the blockchain for transport applications [13]. The use of public blockchain in V2I (vehicle-to-infrastructure) has shown many advantages. However, it also faces significant challenges, especially regarding security. Hackers and fraudsters threaten public, private, and hybrid blockchain networks through phishing, routing, Sybil, and 51% attacks [14]. These attacks exploit known flaws in blockchain technology, leading to hacks and scams. Additional security measures are needed to protect V2I infrastructure, and blockchain can be applied in different layers to enhance security. Phishing attempts target users’ login information by sending fraudulent emails that appear to be from reliable sources. Routing attacks exploit weaknesses in data transfers, intercepting users’ sensitive information [15]. Sybil’s attacks involve creating multiple bogus network identities to overpower the system. To take control of a blockchain network, attackers would need to control over 51% of the mining power. Existing blockchain frameworks like Ethereum, Fabric, and Corda have limitations, particularly regarding cyber-attack efficiency and optimization of energy, time, and delay constraints for vehicle applications in V2I. Addressing these security and optimization challenges will be crucial for realizing the full potential of public blockchain in V2I and ensuring safe and efficient vehicular applications [16]. However, the existing blockchain-enabled V2I architecture still needs improvement, such as in terms of energy, delay, and deadline issues for vehicular applications. (i) The existing studies [2,5,7,8,10] only considered the execution delay in their architectures for vehicular applications. Meanwhile, blockchain suffered from Sybil and 52% attacks and these constraints, deadlines, and delays still degraded the performance of vehicular applications while searching both Sybil and 52%. (ii) These studies consume much more energy and resources while scheduling the tasks of vehicular applications on fog cloud nodes. Therefore, a new architecture should minimize these aforementioned constraints for vehicular applications. This paper presents the V2I paradigm based on cyber-efficient blockchain and scheduling schemes for vehicular applications. The paper formulates the combinatorial problem, where processing time, energy, and malware detection accuracy are constraints for vehicular applications. The paper considers heterogeneous computing nodes such as vehicles, fog, and cloud nodes for data processing. The paper makes the following contribution to the paper: •The study designs a cyber-attack-efficient, secure blockchainenabled V2I paradigm for vehicle applications in cooperative fog and cloud networks. The main novelty is that all existing blockchain frameworks are suffering from cyberattacks during their transactions among nodes. •This paper presents the blockchain schemes that handle cybersecurity attacks such as malware and benign features (e.g., pattern, attribute, signature, and heuristic) on different computing systems. •The adaptive and dynamic consensus methods of blockchain are presented, which determine the dynamic attacks in the V2I paradigm. The paper devises the long short-term memory (LSTM) scheme with the blockchain schemes where different attacks (e.g., phishing, routing, Sybil, and 51%) are identified at the runtime of vehicle applications. •This study presents the scheduling scheme based on the nondominated sorting genetic algorithm-II (NSGA-II) method. The objective is to find the best solution for the problem that involves multiple convex and concave objectives for vehicle applications to different types of virtual machines in a fog and cloud network that works together. The existing NSGA-II [17] has a convergence problem and requires many parameters to select optimal resources and execute all tasks with the given constraints. Therefore, the existing NSGA-II is improved in the proposed work and considered cybersecurity, given the application constraints requirements. The paper is sequenced into different sections. Related work is discussed in Section 2. The proposed and formulated paradigms are presented in Section 3. The paper explains the methodology in Section 4. The performance evaluation is discussed in Section 5. The conclusion and future work are discussed in Section 6. 2. Related work Many important changes have been made to the literature on V2I paradigms. For example, communication, offloading, and scheduling have all been made better for vehicle use. Blockchain technology is being utilized to address security challenges in high-performance V2I computing applications, aiming to decentralize centralized systems. Knowledge-Based Systems 290 (2024) 111576 3 A. Lakhan et al. Table 1 Existing secure blockchain framework in Fog–Cloud-Enabled V2I paradigm. Study App. Node Paradigm Method Objective [1] Vehicle Cloudlet V2I GA Delay [2] Vehicle Cloud V2I LSTM Delay [3] Vehicle VANET V2I NSGA Delay [4] Vehicle Net. V2I MIMO Delay [5–10] Vehicle Cloud VANET Hybrid Security [11–15] Vehicle Cloud V2I Hybrid Many [16,18] Vehicle Hybrid V2I Hybrid Security [19,20] Vehicle Hybrid V2I AI-Hybrid Security Several studies propose blockchain-based frameworks for digital vehicular transport applications, offering efficient validation and execution of heterogeneous services. One such work by Gupta et al. [1] presents V2I and vehicle-to-communication schemes using cloudlet networks to offload and schedule vehicle application workloads with minimal latency. The genetic algorithm (GA) is implemented where cloudlets accept the input as population and apply local searching to find the optimal service for the vehicular applications at the roadside unit. However, these secure blockchain frameworks have the components, research gaps, and architectures discussed in Table 1. The V2I-Carla enabled V2I is presented in [2]. The vehicle dataset is trained based on LSTM, optimizing communication and computation constraints in the cloud network. A vehicular ad-hoc (VANET)-enabled cloud network for vehicular applications is presented in [3]. The objective is to offer distributed services at the roadside unit with minimal delay. The non-dominated sorting genetic algorithm (NSGA) is presented to choose the optimal cloud among the available resources in the network. Tajalli et al. [4] suggested a traffic control-enabled V2I paradigm to improve distributed service network performance, considering only network-related constraints. The multiple input, multiple output (MIMO) scheme is integrated, allowing different networks to take input and generate output on the roadside unit of services. These prior works [5–10] presented centralized security-enabled V2I for vehicular applications based on distributed cloud computing. Hybrid dynamic and adaptive methods are presented to process secure signature data and services among homogeneous and distributed nodes. The vehicle registration security-enabled scheme identifies the valid and secure registration of vehicles in VANET. The role-based authentication scheme assigns secure and valid data processing and downloading roles in distributed services. The Elliptic Curve Cryptographic Algorithm (ECC) converts data into hashing with valid patterns. The malware attacks on V2I were also identified based on their adaptive methods. However, these methods suffered from unknown attacks in their V2I paradigms for vehicle applications. Researchers in the past [11–15] developed dynamic LSTM schemes that let V2I find known and unknown attacks and handle secure data across multiple distributed computing nodes. LSTM was trained and tested on malware datasets for V2I to efficiently identify attacks without degrading application performance. To address various constraints, NSGA-II with Pareto optimal solutions was used to improve energy and delay in V2I. LSTM was employed to identify cyberattacks with benign and malignant symptoms and recognize their patterns based on their families. However, these studies only considered homogeneous nodes and centralized security control in their paradigms. These studies [16,18,21–23] introduced blockchain-enabled V2I paradigms for vehicular applications. To ensure secure vehicle data transaction processing, various blockchain schemes were implemented in fog cloud blocks, including proof of work, creditability, and Byzantine fault tolerance. Public blockchain technologies were utilized in these studies, with smart contract schemes facilitating valid registrations for IoT-enabled vehicle applications. These IoT vehicular applications can offload and download data in an immutable form, even in a heterogeneous environment. Each block in the blockchain has different attributes, such as hashing algorithms (e.g., secure hashing algorithm and advanced standard encryption), transactions, timestamps, and hashing validation schemes. All V2I components, including IoT vehicle devices, base stations (BSS), fog, and cloud nodes, adhere to consensus methods and blockchain validation schemes for secure data processing. However, awareness of new malware attacks remains a key issue in these blockchain-enabled V2I paradigms. These studies [19,20,24–34] suggested artificial intelligence-based blockchain-enabled V2I paradigms for vehicular applications. The goal is to implement blockchain with awareness of known and unknown attacks and ensure secure data transactions in distributed fog cloud networks. The NSGA-II is integrated to optimize the different constraints as convergence functions. However, it consumes much more time and resources and leads to the failure of tasks due to resource scarcity in fog cloud networks. A Sybil assault is a kind of cyberattack in which an adversary gains control of several network nodes and uses them to obstruct regular network functions. Because the blockchain is based on decentralized consensus processes, Sybil’s assaults might seriously jeopardize its security and integrity. A Sybil attack often operates within a blockchain in the following ways: On the network, the attacker generates a huge number of pseudonymous identities or nodes. There could be many nodes in the blockchain that are running independently. The attacker may try to validate contradictory vehicular data transactions on various network segments by taking over a large number of nodes, endangering the blockchain’s integrity. Therefore, blockchain technologies are still suffering from Sybil attacks for vehicular applications. Meanwhile, these studies [35–39] suggested Norwegian air traffic and secure IoT blockchain-assisted mobile edge computing paradigms for healthcare applications that have the same features as vehicular applications. The healthcare services are also deployed at the roadside unit on different edge and cloud networks. IoT devices offload their data to the edge cloud based on blockchain technologies. These studies suggested [40–42] suggested transport application system for shuttle services and obtined the customer data. The sustainble transport is environment presented in these works and obtainted optimal results for transport applications. However, these studies still suffer from different issues. In summary, all existing studies [19,20,24–34] suggested blockchain techniques for vehicular applications and obtained the security advantages for vehicular applications. However, the main limitation with them is not considering the optimal constraints while using blockchain architecture and Sybil attacks are still diverse and consume a lot of resources and time from nodes while processing the vehicle data for execution. Meanwhile, existing V2I paradigms with blockchain for vehicular applications in fog clouds should have considered multi-objective constraints such as processing time, energy, and malware detection accuracy. Therefore, this paper differs from existing studies. This paper shows an effective V2I paradigm that uses blockchain (NSGA-II) and LSTM to solve problems with multiple goals for vehicle applications. 3. Proposed cyber-attacks-efficient secure blockchain V2I paradigm The study presents a cyber-attack-efficient, secure blockchainenabled V2I paradigm consisting of Internet of Things (IoT) vehicular applications, wireless network base stations, and cooperative fog cloud networks, as shown in Fig. 1. The study focuses on coarse-grained vehicular applications, integrating services like trip, fare, train schedule, taxi availability, and others into the applications. Each application features a lightweight blockchain interface with all the necessary properties for data validation. However, considering the public blockchain, the study introduces authentication vehicle registration based on smart contracts to prevent anonymous vehicles from entering the blockchain networks. Because devices only have so many resources, the study comes up with real-time offloading, a way to send all of a device’s work to a cooperative fog cloud through a network of different base Knowledge-Based Systems 290 (2024) 111576 4 A. Lakhan et al. Fig. 1. A multi-objective framework for secure blockchain in the fog–cloud network of vehicle-to-infrastructure application. stations. The roadside unit cooperative fog cloud services form a computing system where various services are deployed at the roadside for vehicular applications. The study considers additional resources in the form of virtual machines. Each fog and cloud offers different virtual machines (e.g., small, medium, large, and extra) with varying speeds and storage capabilities. Depending on the workload, each virtual machine has different energy requirements. This paper utilizes NSGAII to optimize multiple objectives’ concave and convex optimization problems. Specifically, this paper optimizes vehicular IoT applications’ energy, delay, security, and deadline in the blockchain V2I paradigm. The study integrates a secure blockchain within each blockchain, where all rules are deployed in the network. Our blockchain comprises different components, such as workload, hashing, previous hashing, AES-256 [28] random key, long timestamp, fault-tolerant, proof of validation, long short-term memory (LSTM) for cyber-attacks, data training, and modeling for transport applications. However, due to the correlation among different services and tasks, the study develops a multi-learning scheme to minimize processing energy and time for vehicular applications. Table 2 presents the notations used in the problem formulation and algorithms. 3.1. System model and problem formulation This paper assumes that all IoT transport applications, as shown in Fig. 1, are considered vehicle applications. In formulation, the paper considers 𝑁number of vehicle applications as tasks, where 𝑣∈𝑁 is the specific tasks of vehicle application and consists of various coarse-grained services (e.g., trip, fare, location, timetable, and others). Each application has different attributes, such as 𝑣𝑤workload, 𝑣𝑑 deadline 𝑣𝑑, and processing time during scheduling in the system. Table 2 Mathematical notation. Problem notations Notation definitions 𝑁Total numbers of vehicular tasks 𝑣Particular task 𝑑𝑣Deadline of task 𝑤𝑣Workload of task 𝑣𝑠Security requirements of task 𝑤←𝑣The task 𝑣has workload 𝑤 𝐾Number of cooperating fog and cloud node 𝑘Particular node 𝑉 𝑀 Number of virtual machines 𝑣𝑚 ∼𝑡𝑦𝑝𝑒 Type of virtual machine 𝜁𝑣𝑚Speed of virtual machines 𝑃 𝑤𝑣𝑚 Power consumption of vms 𝜖𝑣𝑚 Storage of vm 𝐿Local processing machines 𝜁𝑙Local processing speed 𝐵𝑆 Number of homogeneous base-stations 𝑏𝑠 Particular base-station 𝐵Number of blocks 𝑏Particular block 𝑇Number of timestamps 𝑅𝐾 Number of random key 𝑏−𝑎𝑡𝑡𝑟𝑖𝑏𝑢𝑡𝑒𝑠 Blockchain attributes 𝑀𝑊 [𝑣, 𝑏𝑠, 𝑘, 𝑏]Malware identified in nodes 𝐵𝐸𝑁[𝑣, 𝑏𝑠, 𝑘, 𝑏]Benign Malware in nodes The notation 𝑣←𝑤means that the task 𝑣contains a workload 𝑤. The research considers the 𝐾computing nodes as cooperative fog and cloud networks with varying virtual machine (VM) numbers. All virtual machines (VMs) are different in how fast they work, how much power Knowledge-Based Systems 290 (2024) 111576 5 A. Lakhan et al. they use, and how much storage space they take up on the network. Each computing node has a different number of VMs of various types (small, middle, large, and x-large sizes) to run the vehicular workloads in the system. The study considers the distributed homogeneous base stations and blockchain blocks in the V2I paradigm at roadside units for IoT vehicle applications. The paper denoted the symbols of problem formulation with their descriptions, as shown in Table 2. All the notations have particular values and descriptions, as illustrated in Table 2. 3.2. Cyber-attacks detection and training models This paper presents a robust and resilient LSTM-based scheme to detect both malware and benign at different computing and communication nodes in the V2I paradigm. The malware detection time is determined in the following way: 𝑇 𝑃 = 𝑁 ∑ 𝑣←𝑤=1 ∼𝐵𝑆 𝑏𝑠=1 𝐾 ∑ 𝑘=1 𝐿 ∑ 𝑙=1 𝑤←𝑣∼𝑏𝑠, 𝑙, 𝑘, 𝑏 == 𝑀𝑊 [𝑣, 𝑏𝑠, 𝑘, 𝑏] ∼ 𝐵𝐸𝑁[𝑣, 𝑏𝑠, 𝑘, 𝑏]=1. (1) 𝑇 𝑁 = 𝑁 ∑ 𝑣←𝑤=1 ∼𝐵𝑆 𝑏𝑠=1 𝐾 ∑ 𝑘=1 𝐿 ∑ 𝑙=1 𝑤←𝑣∼𝑏𝑠, 𝑙, 𝑘, 𝑏 == 𝑀𝑊 [𝑣, 𝑏𝑠, 𝑘, 𝑏] ∼ 𝐵𝐸𝑁[𝑣, 𝑏𝑠, 𝑘, 𝑏]=0. (2) Eqs. (1) and (2) determine the malware availability and patterns in the different computing nodes. 3.3. Blockchain attributes and validation The study devises a blockchain mechanism for each node to validate and avoid data from different cyber-attacks in the network. The study determines the blockchain formulation in the following way: 𝑏𝑐𝑚 = 𝑁 ∑ 𝑣←𝑤=1 ∼𝐵 𝑏=1 𝑇 𝑃 +𝑇 𝑁+ 𝐴𝐸𝑆 − 256(𝑣, 𝑤, 𝐿, 𝐾, 𝑟𝑘, 𝑡) 𝜁𝑙 , ∀𝑟𝑘, 𝑡 = 1,…𝑇 , 𝑅𝐾. (3) The blockchain’s validity was checked using Eq. (3) and the current hashing. The generated hashing used the AES-256 encryption scheme and had a unique time stamp, a random public key for validation, and the ability to handle errors for all vehicle applications. 3.4. Offloading and blockchain execution time The study designs the architecture based on a full offloading scheme where all vehicle workloads are offloaded to the cooperative fog cloud for computation. However, a few blockchain operations will be done at the local vehicle before offloading data to the external nodes for processing. The study determined the local processing time in the following ways: 𝐿𝑒 𝑣= 𝑉 ∑ 𝑣=1 𝐿 ∑ 𝑙=1 𝑏𝑐𝑚 ∼𝑣←𝑤 𝜁𝑙 ×𝑥𝑙,𝑣 = 1, ∀𝑣= 1,…, 𝑁. (4) The local blockchain mechanism for verifying vehicle registration and sending work to available cooperative fog and cloud nodes through different BSs is set by Eq. (4). 3.5. V2I deployed BSs in secure blockchain All the IoT vehicle applications in the V2I paradigm are connected to the different BSs on a secure blockchain. Therefore, the vehicle offloads the coarse-grained workload data to the servers via base stations. Therefore, the communication time of the offloading data between nodes is determined in the following way: 𝐶𝑜𝑚 = 𝐿,𝐾,𝐵𝑆 ∑ 𝑙=1,𝑏𝑠=1,𝑘=1 𝑙∼𝑏𝑠 𝑤= 1 𝑏𝑤 + 𝑏𝑠 ∼𝑘𝑓𝑟𝑎𝑐𝑤 = 1𝑏𝑤 +𝑘1 ∼ 𝐾𝑤= 1 𝑏𝑤 +𝑏𝑐𝑚. (5) Eq. (5) determines the communication time between nodes during offloading and processing in the network. 3.6. Remote execution time All the workloads are offloaded to the fog and cloud nodes for processing. Therefore, the study determines the remote execution time in the following way. 𝑅𝑒 𝑣= 𝑉 ∑ 𝑣=1 𝐾 ∑ 𝑘=1 𝑏𝑐𝑚 ∼𝑣←𝑤 𝜁𝑘 ×𝑥𝑙,𝑣 = 1,∀𝑣= 1,…, 𝑁. (6) Eq. (6) determines the remote execution time along with the blockchain validation in work. 3.7. Energy consumption The study determines energy in different ways, such as local processing energy, communication energy, and computing virtual machine energy in the system. The study designs the architecture based on full offloading, where all vehicle workloads are offloaded to the cooperative fog cloud for computation. However, in the following, few blockchain operations will be done at the local vehicle network’s communication time between nodes during offloading and processing local processing time. 𝐿𝐸𝑒 𝑣= 𝑉 ∑ 𝑣=1 𝐿 ∑ 𝑙=1 𝑏𝑐𝑚 ∼𝑣←𝑤 𝜁𝑙 ×𝑥𝑙,𝑣 = 1 × 𝐾𝑊 , ∀𝑣= 1,…, 𝑁. (7) The local energy consumption blockchain mechanism for verifying vehicle registration and sending work to available cooperative fog and cloud nodes through different BSs is found in Eq. (7). All the vehicles are connected to the different BSs. Therefore, the vehicle offloads the coarse-grained workload data to the servers via BSs. Therefore, the communication time of the offloading data between nodes is determined in the following way: 𝐶𝑜𝑚𝐸 = 𝐿,𝐾,𝐵𝑆 ∑ 𝑙=1,𝑏𝑠=1,𝑘=1 𝑙∼𝑏𝑠 𝑤= 1 𝑏𝑤 ×𝐾𝑊 + 𝑏𝑠 ∼𝑘𝑤= 1 𝑏𝑤 ×𝐾𝑊 +𝑘1 ∼ 𝐾𝑓 𝑟𝑎𝑐𝑤 = 1𝑏𝑤 ×𝐾𝑊 +𝑏𝑐𝑚. (8) Eq. (8) determines the communication energy consumption between nodes during offloading and processing in the network. All the workloads are offloaded to the fog and cloud nodes for processing. Therefore, the study determines the remote execution time in the following way: 𝑅𝐸𝑒 𝑣= 𝑉 ∑ 𝑣=1 𝐾 ∑ 𝑘=1 𝑏𝑐𝑚 ∼𝑣←𝑤 𝜁𝑘 ×𝑥𝑙,𝑣 ×𝐾𝑊 = 1, ∀𝑣= 1,…, 𝑁. (9) Eq. (9) determines remote energy consumption along with the blockchain validation in work. Knowledge-Based Systems 290 (2024) 111576 6 A. Lakhan et al. 3.8. Multi-objective function This paper considers multiple objectives for vehicle applications, such as time (e.g., delay), energy, and malware detection accuracy constraints for vehicular applications in distributed computing nodes. The time and energy functions are convex functions to be minimized, whereas accuracy is the concave function to be maximized for all vehicular applications in the system. 𝐹= 𝑁 ∑ 𝑣=1 𝑓1(𝐿𝑒 𝑣) + 𝑓2(𝑅𝑒 𝑣) + 𝑓3(𝐶𝑜𝑚) + 𝑓4(𝑇 𝑃 ), 𝑘= 1,…, 𝐾. (10) The objective is to optimize all objectives for minimizing and maximizing transport applications as defined in Eq. (10). The objective mathematical model is designed in the following way: min 𝐹= 𝑁 ∑ 𝑣=1 𝑓1(𝐿𝑒 𝑣) + 𝑓2(𝑅𝑒 𝑣) + 𝑓3(𝐶𝑜𝑚), 𝑘= 1,…, 𝐾. max 𝑓4(𝑇 𝑃 )𝑘= 1,…, 𝐾. (11) Eq. (11) optimized both convex and concave functions of the vehicular applications on the different nodes. The notation 𝐹shows the multiobjectives, where 𝑘= 1,…, 𝐾 denotes the assignment of all tasks on the distributed fog cloud nodes. 4. Methodology: Non-dominated sorting genetic and LSTM schemes for vehicular applications The study comes up with the adaptive NSGA-II and LSTM methods to solve multi-objective convex and concave problems in the context of vehicles. The proposed adaptive NSGA-II algorithm consists of different methods, as shown in Algorithm 1. The paper does not use convergence and time series prediction of task scheduling and cyber-attack detection in the presented paradigm. The suggested NSGA-II and LSTM schemes work with the blockchain scheme to carry out the task according to the rules set in fog cloud networks. This paper considers the different parameters of the proposed algorithms in terms of energy, time, resources, and security and implements them in all algorithms 1to 6. Algorithm 1: Adaptive NSGA-II: Non-dominated Sorting Genetic Algorithm Input : 𝑉 , 𝐾, 𝐵, 𝐵𝑆 1begin 2foreach (V as v) do 3Initial Population Scheme ; 4Evaluate Individual Fitness Scheme ; 5Select Ranked Resources scheme; 6Select Non-Dominated Sorting Resources; 7Multi-Task Learning Scheme ; 8Multiple Patterns Enabled LSTM scheme; 9End process; All the inputs of vehicular application workloads randomly arrive at the system and need services and resources for execution in V2I. The combinatorial problem has many conflicting objectives for the minimum and maximum functions for vehicular applications. Algorithm 1 proposed the adaptive solution and consisted of different schemes for vehicular applications. 4.1. Initial population scheme The paper presents a secure blockchain scheme to process the inputs to prevent cyber-attacks. The algorithm initially takes the vehicular workloads as input, represented as 𝑣= 1,…, 𝑉 in the population. Fig. 2. Secure blockchain-enabled input from vehicular applications. These workloads are coarse-grained and fully offloaded to the fog and cloud servers. To make sure that all vehicular apps are legitimate, they are registered using smart contracts for authentication. Their plaintext data is then hashed using AES-256 [28] and various timestamps, as seen in Fig. 2. The workloads are sorted based on their deadlines in sequential order and then scheduled individually for execution in parallel computing processing. Algorithm 2: Initial Population Scheme Input : 𝑉 , 𝑣𝑤 1begin 2Offload[𝑣, 𝑣𝑤, 𝑙]; 3foreach (𝑉as 𝑣𝑤)do 4if (𝑣, 𝑤, 𝑙)then 5Determined the local execution time based on equation (4); 6𝑏𝑐𝑚 =∑𝑁 𝑣←𝑤=1 ∼𝐵 𝑏=1 𝑇 𝑃 +𝑇 𝑁 +𝐴𝐸𝑆−256(𝑣,𝑤,𝐿,𝐾,𝑟𝑘,𝑡) 𝜁𝑙 ; 7Offload[𝑣, 𝑣𝑤, 𝑙]=∑𝑁 𝑣←𝑤=1 ∼𝐵𝑆 𝑏𝑠=1 ∑𝐾 𝑘=1 ∑𝐿 𝑙=1; 8𝑤←𝑣∼𝑏𝑠, 𝑙, 𝑘, 𝑏 == 𝑀𝑊 [] ∼ 𝐵𝐸𝑁[] = 1; 9Determined the secure blockchain processing based on equation (1) and equation(2); 10 Offload[𝑣, 𝑣𝑤, 𝑙]=𝑇 𝑃 =∑𝑁 𝑣←𝑤=1 ∼𝐵𝑆 𝑏𝑠=1 ∑𝐾 𝑘=1 ∑𝐿 𝑙=1; 11 𝑤←𝑣∼𝑏𝑠, 𝑙, 𝑘, 𝑏 == 𝑀𝑊 [] ∼ 𝐵𝐸𝑁[] = 1; 12 Offload[𝑣, 𝑣𝑤, 𝑙] ∼ 𝑀𝑊 [𝑂𝑓𝑓 𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙] ∼ 𝐵𝐸𝑁[𝑂𝑓𝑓 𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙]; 13 Call Evaluate Individual Fitness Scheme; 14 End process; Algorithm 2has the following actions for the given vehicular workloads: Knowledge-Based Systems 290 (2024) 111576 7 A. Lakhan et al. •The vehicular applications input the data from scratch in the system as defined in 𝑏𝑐𝑚 =∑𝑁 𝑣←𝑤=1 ∼𝐵 𝑏=1 𝑇 𝑃 +𝑇 𝑁 + 𝐴𝐸𝑆−256(𝑣,𝑤,𝐿,𝐾,𝑟𝑘,𝑡) 𝜁𝑙 . All the vehicular applications registered in the system are based on smart contract rules, meaning vehicle authentication is necessary in the framework. •The blockchain processing must be applied before offloading to the fog and cloud for processing. Offload[𝑣, 𝑣𝑤, 𝑙] = ∑𝑁 𝑣←𝑤=1 ∼𝐵𝑆 𝑏𝑠=1 ∑𝐾 𝑘=1 ∑𝐿 𝑙=1. It means all the data is offloaded after the authentication and blockchain processes in the framework. All the malware and benign attacks are verified during offloading in the system as defined in the rules, e.g., 𝑤←𝑣∼𝑏𝑠, 𝑙, 𝑘, 𝑏 == 𝑀𝑊 [𝑣, 𝑏𝑠, 𝑘, 𝑏] ∼ 𝐵𝐸𝑁[𝑣, 𝑏𝑠, 𝑘, 𝑏]=1. 4.2. Evaluate individual fitness scheme Once the initial authentication and blockchain process is done, Algorithm 3checks the workloads’ fitness using blockchain hashing, proof of work, and attacks to see which ones are good and which ones are bad. In the blockchain blocks, the study implemented four different schemes, namely pattern, attributes, heuristics, and signatures, to predict and verify the presence of malware and benign data inside the blocks before offloading them to the fog and cloud for execution. As shown in Fig. 3, all validations in secure blockchain Algorithm 3: Evaluate Individual Fitness scheme Input : Offload[𝑣, 𝑣𝑤, 𝑙] ∼ 𝑀𝑊 [𝑂𝑓𝑓 𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙] ∼ 𝐵𝐸𝑁[𝑂𝑓𝑓 𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙]; 1begin 2Offload[𝑣, 𝑣𝑤, 𝑙]; 3foreach (𝑉as 𝑣𝑤)do 4if (𝑣, 𝑤, 𝑙 ∼𝑀𝑊 [𝑂𝑓 𝑓𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙]==true) then 5Call appropriate schemes Heuristic, Pattern, signature, Attributes; 6End process; validation are done by algorithm 3using different methods. This is done after all vehicle workloads are authenticated and checked in the blockchain architecture. The security in the blockchain is validated based on malware and benign malware detection, including phishing, routing, Sybil, 51%, and others in the system. All workloads, denoted as 𝑣= 1,…, 𝑁, are converted into hashing using AES-256, and their hashing must be clean from any attacks, such as benign and malware attacks in the system, as shown in Fig. 3. 4.3. Select ranked resources scheme This study considers different computing nodes, such as fog and cloud nodes, with different virtual machines (VMs) with distinct speeds and resources. Algorithm 4determines available virtual machines with different types when they have more resources than the requested workloads for executions. Various virtual machines exist in different computing nodes, such as small, middle, large, and x-large. Algorithm 4identifies the available VMs and enlists those that are not available and partially available in the network, as shown in Fig. 4. The light blue nodes indicate that the resource VMs with particular types (small, middle, large, and x-large) are unavailable for the requested workloads and must search for other nodes. On the other hand, the green nodes show that VMs are available, and the yellow ones show that the VMs are available, but they have some resources that are not sufficient for execution. 4.4. Select non-dominated sorting resources based on task learning The study proposes a multi-tasking, efficient learning scheme in which multiple tasks can be completed concurrently. The main idea is to combine tasks with similar characteristics regarding execution, Fig. 3. Secure blockchain local validation and authentication. Algorithm 4: Select Ranked Resources scheme Input : 𝑁, 𝑘 = 1,…, 𝐾; 1begin 2foreach (𝑘as 𝐾)do 3𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒[𝑉 𝑀𝑠, 𝑇 𝑦𝑝𝑒, 𝑘]Determined the resource availability of nodes based on 𝜖𝑘; 4if (𝑉 𝑀 ←𝑇 𝑦𝑝𝑒 ←𝑘1𝜖𝑘≤𝑘1)then 5Make the VMs status available; 6𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒[𝑉 𝑀𝑠, 𝑇 𝑦𝑝𝑒, 𝑘]=𝑅𝑒𝑠𝑜𝑢𝑟𝑐𝑒[𝑉 𝑀𝑠, 𝑇 𝑦𝑝𝑒, 𝑘]=1; 7else if (𝑉 𝑀 ←𝑇 𝑦𝑝𝑒 ←𝑘1𝜖𝑘≤𝑘1..𝐾)then 8Make the VMs status Unavailable; 9𝑉 𝑀 ←𝑇 𝑦𝑝𝑒 ←𝑘1𝜖𝑘≤𝑘1..𝐾 = 0 10 else 11 Make the VMs status Partial Available; 12 End Conditions; 13 End Searching the availability of VMs; 14 End process; energy, malware, and benign detection during network scheduling. The non-dominated sorting resource-enabled multi-tasking scheme ensures that all requested workloads are executed in parallel, sharing standard features and service results. For example, services like fare, trip, train, bus, and location can run simultaneously and use different processing methods. However, the taxi services requested show that the resource VMs with particular types (small, middle, large, and x-large) are not available for the requested workloads, and thus, they must search for other available options, which are different from the green nodes that indicate availability and the yellow nodes that indicate partially Knowledge-Based Systems 290 (2024) 111576 8 A. Lakhan et al. Fig. 4. Resource ranks based on availability of VMs. available VMs. Algorithm 5determines the parallel execution of Algorithm 5: Select Nominated Sorting Resources Based Multi-Tasks Learning Scheme Input : 𝑘= 1,…, 𝐾, 𝑀𝑊 [𝑂𝑓𝑓𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙] ∼ 𝐵𝐸𝑁[𝑂𝑓𝑓 𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙]; 1begin 2Offload[𝑣, 𝑣𝑤, 𝑙]; 3Initialize Common[] variable; 4foreach (𝑉as 𝑣𝑤)do 5if (𝑣, 𝑤, 𝑙 ∼𝑀𝑊 [𝑂𝑓 𝑓𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙]==true) then 6Determine cyber-attack attributes Heuristic, Pattern, signature, Attributes; 7Determined the features in first step; 8𝑀𝑊 [𝑂𝑓𝑓𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙] ∼ 𝐵𝐸𝑁[𝑂𝑓𝑓 𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙]← Attributes; 9Common∼Cluster[]𝑀𝑊 [𝑂𝑓𝑓𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙] ∼ 𝐵𝐸𝑁[𝑂𝑓𝑓 𝑙𝑜𝑎𝑑[𝑣, 𝑣𝑤, 𝑙]; 10 Schedule Parallel 𝑣, 𝑣𝑤, 𝑘 on different VMs among 𝐾 with lower energy and time; 11 End Multi-Tasking Process; 12 End Clustering; 13 End process; computing workloads and identifies attacks with different schemes. For instance, each malware and benign attack is identified based on four methods: pattern, heuristic, signature, and attributes. The main reason for using these schemes is that all attacks have different forms and may have a family relationship. Hence, multiple identifier schemes are needed to detect malware and benign data, covering all attacking techniques in the V2I paradigm. Fig. 5 illustrates the multi-task learning capability with different transport workloads for the execution of the same computing services. It shows that parallel and similar services for workloads are executed concurrently with the same workload services in the network. All attacks with different schemes can be identified at one node and shared with others if they require similar services. We select nominated sorting resources based on a multi-tasking learning scheme; therefore, we schedule similar workloads on the same types of VMs. Furthermore, this paper determined attacks with the same pattern and shared their copies with another node to minimize resource, energy, and time consumption. Therefore, execution results of 𝑣5are copied to parallel workloads 𝑣6,…, 𝑁, as shown in Fig. 5. We determined all attacks based on heuristics, pattern, signature, and attributes on different VMs with their types in different nodes. The determined attacks and types try to recover from scheduling workloads on different nodes with the minimum energy consumption and time. Therefore, we search for the few available VMs among the nodes. 4.5. Multiple patterns enabled LSTM scheme for scheduling In the suggested V2I paradigm, the paper considers heterogeneous computing nodes where vehicular applications request services and offload their data onto different computing nodes. The blockchainenabled computing nodes can securely process the application data and identify cyber-attacks within the V2I paradigm. It is essential to execute all applications under their given constraints to account for past and new attacks in the sequencing along with the applications. We devise an LSTM-enabled scheduling scheme in which energy, time, prediction of attacks, and security constraints are optimized for the requested services in V2I. The ratio of cyber-attacks in the blockchain-based V2I also exhibits diversity; new attacks, such as Sybil’s 51% of attacks, degrade the overall performance of the blockchain for applications. The existing cuckoo sandbox based on LSTM [29] only works well when multiple objectives and constraints are associated with applications. Therefore, we suggested the LSTM cyber-attack detection in blockchain-based V2I The paper defined all the steps of Algorithm 6in the following way: •All computing nodes offload and process vehicle applications based on the blockchain scheme. We scan the diversity of cyberattacks based on LSTM with the forget, input, and output gates. We clustered the same types of cyberattacks into a list based on past data and classifications. •Steps 1 to 4 show that each computing node must process the data based on blockchain technology. All the blockchain blocks, e.g., 𝑏=1∈𝐵, are initialized based on their attributes. •In the algorithm, steps 5–12 describe how the vehicle workloads are handled on their local computers. The data transaction is based on blockchain technology, and if cyberattacks happen and the status changes to true, an LSTM-based blockchain scheme finds the malware features. The algorithm implemented four cyber-attack schemes: heuristic, hashing, pattern, and signature, and searched for the optimal recovery solution based on Algorithm 5. These cyber-attacks are part of the LSTM classification with the positive true ratio, where all attacks are listed based on their features. •The Algorithm determined the local processing time, communication time, and energy during execution and offloading from local computing nodes to fog cloud computing with the security validations. •The offloaded workloads are scheduled on different computing nodes, as defined in steps 11–18. •Based on NGSA-II, the algorithm optimized all the objectives of the vehicular applications based on their constraints. •Algorithm optimize both concave and convex objectives with maxima and minima as shown in these functions: max 𝑓4(𝑇 𝑃 ) 𝑘= 1,…, 𝐾 and max 𝑓4(𝑇 𝑃 )𝑘= 1,…, 𝐾. The VMs are essential, where multi-tasking schemes search similar VMs in different computing nodes. In other aspects, the scheduler, as shown in Algorithm 6, determined the cyber-attacks VMs in blockchain and tried to schedule executing workloads to other VMs with the minimum power consumption, time, and resources. In this case, the scheduler must search the small VMS according to the required resources for workloads and their deadlines. Knowledge-Based Systems 290 (2024) 111576 9 A. Lakhan et al. Fig. 5. Multi-tasking and LSTM-enabled scheduling. Fig. 5 shows that each fog and cloud node has different virtual machines shown via different colors. Each fog node has four virtual machine types: small, middle, large, and x-large. All the fog and clouds are cooperative and can share their data and migrate their workloads during submission to the network. There are two types of functions, concave and convex, with tightness regarding optimal-function point [1,...,10] thresholds in the V2I paradigm. You can see that algorithm 6is flexible and improves all objective functions in Figs. 6 and 7. It does this for both concave and convex functions for all scheduled workloads. Fig. 6 shows the convex optimization of the functions, where the 𝑦axis represents the functions and the 𝑥-axis represents the function improvement in terms of a given threshold. Where functions have lower to higher delays, as shown in points 0 to 100. Algorithm 6minimizes the delays gradually by achieving the virtual machines and minimizing the energy consumption in the V2I paradigm. Fig. 7 optimizes the concave optimization function of the objective function 𝐹where the 𝑦-axis shows the maximized performance during malware and benign identification in distributed resources of the V2I paradigm. 5. Performance evaluation In this session, the study designs the practical and real-case studies based on simulation for the problem. The simulation comprises distributed fog and cloud services on the road-unit side, resource allocation, cyber-attacks, and multi-objective schemes for vehicular applications. The simulator, as shown in Table 3, consists of different parameters that are integrated into the simulation setting for execution. Fig. 6. Optimization in conflicting convex objectives in terms of functions. The services are deployed at the road unit-side, and the particular coordinates, locations, resolutions, and transport speeds are shown in Fig. 8 and Table 3. The application services are integrated with the IoT vehicular workload of applications determined in megabytes (MB). Table 4 defines the workloads with their descriptions. At the same time, 𝑁shows the total tasks, e.g., 𝑁= 400.Table 5 shows the configured resources of the V2I and simulation for vehicular applications. In this study, the paper designed a malware dataset with known and unknown malware lists, as shown in Table 6. The different kinds of Knowledge-Based Systems 290 (2024) 111576 16 A. Lakhan et al. acquisition, Writing and editing the manuscript. Radek Martinek: Project administration, Funding acquisition, Writing and editing the manuscript. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Data availability The paper exploited the distributed scenario of the V2I paradigm based on the available scenario, as shown on this URL: EveryWareLab scenario:(everywarelab.di.unimi.it/index.php/lbs-datasim). The application workloads are available on the following URL: https://github. com/ABDULLAH-RAZA/Blockchain-Socket. The malware dataset is available on the following link: https://github.com/ABDULLAH-RAZA/ Assignment-/blob/master/Blockchain-Malware-Detection.csv. References [1] M. Gupta, J. Benson, F. Patwa, R. Sandhu, Secure V2V and V2I communication in intelligent transportation using cloudlets, IEEE Trans. Serv. Comput. 15 (4) (2020) 1912–1925. [2] H. Wang, X. Yuan, Y. Cai, L. Chen, Y. Li, V2i-carla: a novel dataset and a method for vehicle reidentification-based v2i environment, IEEE Trans. Instrum. Meas. 71 (2022) 1–9. [3] M. Soni, B.S. Rajput, T. Patel, N. Parmar, Lightweight vehicle-to-infrastructure message verification method for VANET, in: Data Science and Intelligent Applications, Springer, 2021, pp. 451–456. [4] M. Tajalli, M. Mehrabipour, A. Hajbabaie, Network-level coordinated speed optimization and traffic light control for connected and automated vehicles, IEEE Trans. Intell. Transp. Syst. 22 (11) (2020) 6748–6759. [5] M. Umar, S.H. Islam, K. Mahmood, S. Ahmed, Z. Ghaffar, M.A. Saleem, Provable secure identity-based anonymous and privacy-preserving inter-vehicular authentication protocol for VANETS using PUF, IEEE Trans. Veh. Technol. 70 (11) (2021) 12158–12167. [6] L. Wei, J. Cui, H. Zhong, Y. Xu, L. Liu, Proven secure tree-based authenticated key agreement for securing V2V and V2I communications in VANETs, IEEE Trans. Mob. Comput. 21 (9) (2021) 3280–3297. [7] I. Ali, Y. Chen, N. Ullah, R. Kumar, W. He, An efficient and provably secure ECC-based conditional privacy-preserving authentication for vehicle-to-vehicle communication in VANETs, IEEE Trans. Veh. Technol. 70 (2) (2021) 1278–1291. [8] Z. Hassan, A. Mehmood, C. Maple, M.A. Khan, A. Aldegheishem, Intelligent detection of black hole attacks for secure communication in autonomous and connected vehicles, IEEE Access 8 (2020) 199618–199628. [9] G. Thumbur, G.S. Rao, P.V. Reddy, N. Gayathri, D.K. Reddy, M. Padmavathamma, Efficient and secure certificateless aggregate signature-based authentication scheme for vehicular ad hoc networks, IEEE Internet Things J. 8 (3) (2020) 1908–1920. [10] J. Yu, M. Gao, Y. Li, Z. Zhang, W.H. Ip, K.L. Yung, Workflow performance prediction based on graph structure aware deep attention neural network, J. Ind. Inf. Integr. 27 (2022) 100337. [11] M. Altahrawi, N.F. Abdullah, R. Nordin, Service-oriented LSTM multi-criteria RAT selection scheme for vehicle-to-infrastructure communication, IEEE Access 10 (2022) 110261–110284. [12] P. Bhadauria, R. Kumar, S. Sharma, Beamforming in vehicle to infrastructure scenario with respect to LSTM and NAR method, J. Electr. Eng. Technol. (2023) 1–14. [13] Y. Ni, X. Li, H. Zhao, J. Yang, W. Xia, G. Gui, An effective hybrid V2V/V2I transmission latency method based on LSTM neural network, Phys. Commun. 51 (2022) 101562. [14] C. Cintrano, J. Toutouh, Multiobjective electric vehicle charging station locations in a city scale area: malaga study case, in: International Conference on the Applications of Evolutionary Computation, Part of EvoStar, Springer, 2022, pp. 584–600. [15] X. Lin, Z. Lin, S. Wei, Multi-objective optimized driving strategy of dualmotor EVs using NSGA-II as a case study and comparison of various intelligent algorithms, Appl. Soft Comput. 111 (2021) 107684. [16] Z. Yang, K. Yang, L. Lei, K. Zheng, V.C. Leung, Blockchain-based decentralized trust management in vehicular networks, IEEE Internet Things J. 6 (2) (2018) 1495–1505. [17] M. Yuan, Y. Li, L. Zhang, F. Pei, Research on intelligent workshop resource scheduling method based on improved NSGA-II algorithm, Robot. Comput.-Integr. Manuf. 71 (2021) 102141. [18] M.U. Javed, M. Rehman, N. Javaid, A. Aldegheishem, N. Alrajeh, M. Tahir, Blockchain-based secure data storage for distributed vehicular networks, Appl. Sci. 10 (6) (2020) 2011. [19] M. Wazid, B. Bera, A.K. Das, S.P. Mohanty, M. Jo, Fortifying smart transportation security through public blockchain, IEEE Internet Things J. 9 (17) (2022) 16532–16545. [20] E. Rabieinejad, A. Yazdinejad, A. Dehghantanha, R.M. Parizi, G. Srivastava, Secure ai and blockchain-enabled framework in smart vehicular networks, in: 2021 IEEE Globecom Workshops, GC Wkshps, IEEE, 2021, pp. 1–6. [21] J. Guo, X. Ding, W. Wu, Reliable traffic monitoring mechanisms based on blockchain in vehicular networks, IEEE Trans. Reliab. (2021). [22] C. Zhang, W. Li, Y. Luo, Y. Hu, AIT: An AI-enabled trust management system for vehicular networks using blockchain technology, IEEE Internet Things J. 8 (5) (2020) 3157–3169. [23] Y. Qu, L. Gao, Y. Xiang, S. Shen, S. Yu, FedTwin: Blockchain-enabled adaptive asynchronous federated learning for digital twin networks, IEEE Netw. (2022). [24] L. Benarous, B. Kadri, A. Bouridane, Blockchain-based privacy-aware pseudonym management framework for vehicular networks, Arab. J. Sci. Eng. 45 (8) (2020) 6033–6049. [25] Z. Yang, R. Yang, F.R. Yu, M. Li, Y. Zhang, Y. Teng, Sharded blockchain for collaborative computing in the internet of things: Combined of dynamic clustering and deep reinforcement learning approach, IEEE Internet Things J. (2022). [26] C. Zhang, W. Li, Y. Luo, Y. Hu, AIT: An AI-enabled trust management system for vehicular networks using blockchain technology, IEEE Internet Things J. 8 (5) (2020) 3157–3169. [27] S.K. Rana, S.K. Rana, K. Nisar, A.A. Ag Ibrahim, A.K. Rana, N. Goyal, P. Chawla, Blockchain technology and artificial intelligence based decentralized access control model to enable secure interoperability for healthcare, Sustainability 14 (15) (2022) 9471. [28] S.J. Gabriel, P. Sengottuvelan, An enhanced blockchain technology with AES encryption security system for healthcare system, in: 2021 2nd International Conference on Smart Electronics and Communication, ICOSEC, IEEE, 2021, pp. 400–405. [29] S. Rana, N. Kumar, A. Handa, S.K. Shukla, Automated windows behavioral tracing for malware analysis, Secur. Privacy 5 (6) (2022) e253. [30] S. Zhu, H. Chen, H. Zhu, R. Qiao, et al., ‘‘Improved Nsga-Ii Algorithm-Based Task Offloading Decision in Internet of Vehicles Edge Computing Scenario’’, Hao and Zhu, Hai and Qiao, Rui, Improved Nsga-Ii Algorithm-Based Task Offloading Decision in Internet of Vehicles Edge Computing Scenario. [31] M. Su, C. Cao, M. Dai, J. Li, Y. Li, Towards fast and energy-efficient offloading for vehicular edge computing, in: 2022 IEEE 28th International Conference on Parallel and Distributed Systems, ICPADS, IEEE, 2023, pp. 649–656. [32] T. Rajabi, A.A. Khalil, M.H. Manshaei, M.A. Rahman, M. Dakhilalian, M. Ngouen, M. Jadliwala, A.S. Uluagac, Feasibility analysis for sybil attacks in shard-based permissionless blockchains, Distributed Ledger Technol.: Res. Pract. 2 (4) (2023) 1–21. [33] A. Hafid, A.S. Hafid, M. Samih, A tractable probabilistic approach to analyze sybil attacks in sharding-based blockchain protocols, IEEE Trans. Emerg. Top. Comput. 11 (1) (2022) 126–136. [34] S. Benadla, O.R. Merad-Boudia, S.M. Senouci, M. Lehsaini, Detecting sybil attacks in vehicular fog networks using RSSI and blockchain, IEEE Trans. Netw. Serv. Manag. 19 (4) (2022) 3919–3935. [35] J.A. Alzubi, O.A. Alzubi, A. Singh, T. Mahmod Alzubi, A blockchain-enabled security management framework for mobile edge computing, Int. J. Netw. Manag. 33 (5) (2023) e2240. [36] O.A. Alzubi, J.A. Alzubi, K. Shankar, D. Gupta, Blockchain and artificial intelligence enabled privacy-preserving medical data transmission in internet of things, Trans. Emerg. Telecommun. Technol. 32 (12) (2021) e4360. [37] J.A. Alzubi, O.A. Alzubi, A. Singh, M. Ramachandran, Cloud-IIoT-based electronic health record privacy-preserving by CNN and blockchain-enabled federated learning, IEEE Trans. Ind. Inform. 19 (1) (2022) 1080–1087. [38] S. Yassine, A. Stanulov, A comparative analysis of machine learning algorithms for the purpose of predicting norwegian air passenger traffic, Int. J. Math. Statist. Comput. Sci. 2 (2024) 28–43. [39] H.C. Altunay, Z. Albayrak, A hybrid CNN+ LSTMbased intrusion detection system for industrial IoT networks, Eng. Sci. Technol. Int. J. 38 (2023) 101322. [40] S. Moslem, H. Solieman, L. Oubahman, S. Duleba, T. Senapati, F. Pilla, Assessing public transport supply quality: a comparative analysis of analytical network process and analytical hierarchy process, J. Soft Comput. Decis. Anal. 1 (1) (2023) 124–138. [41] K.A. Tütüncü, N.N. Gül, U. Bölükbaş, A.F. Güneri, Integer linear programming approach for the personnel shuttles routing problem in yıldız campus in istanbul, J. Soft Comput. Decis. Anal. 1 (1) (2023) 303–316. [42] H.K.Y. Al-Zibaree, M. Konur, Fuzzy analytic hierarchal process for sustainable public transport system, J. Oper. Intell. 1 (1) (2023) 1–10.