No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
A Federated Transfer Learning-Empowered Blockchain-Enabled Secure Knowledge Sharing Scheme for Unmanned Any Vehicles in Smart Cities
Abstract
Smart cities embrace unmanned autonomous vehicles (UxVs) for urban mobility and addressing challenges. UxVs include UAVs, UGVs, USVs, and UUVs, empowered by AI, particularly deep learning (DL), for autonomous missions. However, traditional DL has limitations in adapting to dynamic environments and raises data privacy concerns. Limited data availability and starting from scratch to adapt to a new environment during missions pose challenges. Additionally, cyber threats, particularly in terms of communication and data security, can jeopardize the missions performed by UxVs. This paper proposes a federated transfer learning scheme for UxVs, sharing prior knowledge and training with limited data while ensuring security through blockchain. Domain adaptation with maximum mean discrepancy enhances the DL model's performance in target domains. The proposed scheme's feasibility is demonstrated in an empirical environment, and it outperforms existing works.
... The integration of unmanned any vehicles (UxVs), encompassing unmanned aerial vehicles (UAVs), unmanned ground vehicles (UGVs), unmanned surface vehicles (USVs), and unmanned underwater vehicles (UUVs), is transforming smart ecosystems by enhancing critical infrastructure, industries, and smart cities [1]. Equipped with advanced sensors and autonomous navigation, these vehicles provide notable benefits such as increased efficiency, improved safety, and cost savings [2]. ...
Unmanned any vehicles (UxVs), including aerial (UAVs), ground (UGVs), surface (USVs), and underwater (UUVs) vehicles, are transforming smart ecosystems by enhancing infrastructure, industry, and urban environments. However, the diverse technologies involved in UxVs make standardization and efficient management in ultra-dense networks challenging. Zero Touch Network (ZTN) technology leverages artificial intelligence and software-defined networks to automate network management, addressing these complexities but also introducing vulnerabilities to cyber threats such as False Data Injection (FDI) attacks. This paper introduces a secure data consolidation scheme where UxV operation data is validated with a deep learning model before being integrated into the ZTN for seamless management. A simplified transformer-based FDI attack detection model is proposed, enhanced through data augmentation and information gain-based feature selection. An experimental environment is established to demonstrate the feasibility of the proposed scheme, and it is shown that the proposed scheme outperforms existing methods.
This paper proposes a blockchain-based firmware update method using unmanned aerial vehicles (UAVs) to solve one of the security issues arising in the Internet of Things (IoT) environment, which is the firmware security problem. It has high scalability and transaction speed using private blockchains and solves the limitations of internet connections by updating the firmware using an UAV. The proposed firmware update system safely manages the IoT device and firmware information through four processes: participant registration, firmware registration/update, firmware update request, and firmware update. The verification of IoT devices and UAVs is performed using the IoT device’s public key and Bloom filter, and firmware updates can be safely performed using public-key encryption communication. To prove the security of the proposed method, a security analysis based on the STRIDE model was conducted, and the performance of the blockchain network was analyzed by simulation on the Hyperledger.
The prevalence of heart disease has remained a major cause of mortalities across the world and has been challenging for healthcare providers to detect early symptoms of cardiac patients. To this end, several conventional machine learning models have gained popularity in providing precise prediction of heart diseases by taking into account the underlying conditions of patients. The drawbacks associated with these methods are a lack of generalization and the convergence rate of these methods being much slower. As the healthcare data associated with these systems scale up leading to healthcare big data issues, a Cloud-Fog computing-based paradigm is necessary to facilitate low-latency and energy-efficient computation of the healthcare data. In this paper, a DeepMist framework is suggested which exploits Deep Learning models operating over Mist Computing infrastructure to leverage fast predictive convergence, low-latency, and energy efficiency for smart healthcare systems. We exploit the Deep Q Network (DQN) algorithm for building the prediction model for identifying heart diseases over the Mist computing layer. Different performance evaluation metrics, like precision, recall, f-measure, accuracy, energy consumption, and delay, are used to assess the proposed DeepMist framework. It provided an overall prediction accuracy of 97.6714 % and loss value of 0.3841, along with energy consumption and delay of 32.1002 mJ and 2.8002 ms respectively. To validate the efficacy of DeepMist, we compare its outcomes over the heart disease dataset in convergence with other benchmark models like Q-Reinforcement Learning (QRL) and Deep Reinforcement Learning (DRL) algorithms and observe that the proposed scheme outperforms all others.
The latest 5G mobile networks have enabled many exciting Internet of Things (IoT) applications that employ Un-manned Aerial Vehicles (UAVs/drones). The success of most UAV-based IoT applications is heavily dependent on artificial intelligence (AI) technologies, for instance, computer vision and path planning. These AI methods must process data and provide decisions while ensuring low latency and low energy consumption. However, the existing cloud-based AI paradigm finds it difficult to meet these strict UAV requirements. Edge AI, which runs AI on-device or on edge servers close to users, can be suitable for improving UAV-based IoT services. This paper provides a comprehensive analysis of the impact of edge AI on key UAV technical aspects (i.e., autonomous navigation, formation control, power management, security and privacy, computer vision, and communication) and applications (i.e., delivery systems, civil infrastructure inspection, precision agriculture, search and rescue operations, acting as aerial wireless BSs and drone light shows). As guidance for researchers and practitioners, the paper also explores UAV-based edge AI implementation challenges, lessons learned, and future research directions.
This paper presents a deep learning (DL) model integrated automated and secure garbage management scheme using unmanned any vehicle (UxV) to minimize the human effort in terms of the traditional garbage management system. Different kinds of UxV (unmanned aerial vehicles, automated guided vehicles, unmanned surface vehicles, unmanned underwater vehicles, etc.) are utilized to establish an automated garbage management scheme to collect and place the garbage both from the ground and sea surfaces. However, due to the limited battery capacity and inadequate resources of different UxV, a lightweight DL model is developed to detect the garbage successfully with a higher accuracy rate. The proposed lightweight DL model uses two activation functions named MISH and rectified linear unit to enhance the feature extraction and detects the garbage. Moreover, a multi-access edge computing (MEC) server is allocated in the proposed scheme to improve the quality of service (QoS) (i.e., reduce latency and improve security). Furthermore, a blockchain-based secure hazardous garbage (e.g., infectious, toxic, or radioactive materials) tracking technique is concluded in this scheme to identify the individual and reduce the potential harm to the environment. Experimental results demonstrate that the UxV can successfully detect the garbage using the proposed lightweight DL model within a minimum time frame and the obtained accuracy is higher than the other existing DL models. Besides, QoS analysis has been performed to verify the efficacy of the proposed scheme. Finally, a private blockchain network is established to demonstrate the performance of the proposed hazardous garbage tracking technique.
In recent years, there has been an enormous interest in using deep learning to classify underwater images to identify various objects like fishes, plankton, coral reefs, seagrass, submarines, and gestures of sea-divers. This classification is essential for measuring the water bodies' health and quality and protecting the endangered species. Further, it has applications in oceanography, marine economy and defense, environment protection, and underwater exploration and human-robot col-laborative tasks. This paper presents a survey of deep learning techniques for performing the underwater image classification. We underscore the similarities and differences of several methods. We believe that underwater image classification is one of the killer application that would test the ultimate success of deep learning techniques. Towards realizing that goal, this survey seeks to inform researchers about state-of-the-art on deep learning on underwater images and also motivate them to push its frontiers forward. Index Terms-Deep neural networks, artificial intelligence, autonomous underwater vehicle, transfer learning.
Federated learning (FL) allows UAVs to collaboratively train a globally shared machine learning model while locally preserving their private data. Recently, the FL in edge-aided unmanned aerial vehicle (UAV) networks has drawn an upsurge of research interest due to a bursting increase in heterogeneous data acquired by UAVs and the need to build the global model with privacy; however, a critical issue is how to deal with the non-independent and identically distributed (non-i.i.d.) nature of heterogeneous data while ensuring the convergence of learning. To effectively address this challenging issue, this paper proposes a novel and high-performing FL scheme, namely, the hierarchical FL algorithm, for the edge-aided UAV network, which exploits the edge servers located in base stations as intermediate aggregators with employing commonly shared data. Experiment results demonstrate that the proposed hierarchical FL algorithm outperforms several baseline FL algorithms and exhibits better convergence behavior.
The Internet of Things (IoT) offers new services in the context of smart cities through digital devices embedded with sensing, computation, and communication capabilities. The IoT devices enhance the smart city vision by employing advanced communication and computation technologies for smart city administrations. The IoT-based smart city applications require many IoT devices and gateways to be deployed at different city points. Heterogeneous sensing devices, placing smart devices in a constrained or physically inaccessible area, and large urban areas to monitor together make IoT node deployment and sensing management tasks difficult, time-consuming, and expensive. Additionally, certain tasks may require smart devices to be deployed for a very short period of time to sense and report contextual information, making it economically infeasible to purchase the devices. In this regard, we propose a drone-based IoT as a service (IoTaaS) framework that enables the dynamic provisioning or deployment of IoT devices using drones. IoTaaS allows IoT devices and gateways to be mounted on drones and provides a distributed cloud service by placing the IoT devices in an area according to the requirements specified by a user. We also provide an economic analysis for operating such dronebased IoT services. A proof-of-concept implementation of IoTaaS for smart agriculture and air pollution monitoring applications shows that IoTaaS can reduce setup costs and increase the usage of IoT devices.
Contagious disease pandemics present a significant threat worldwide in terms of both human health and economic damage. New diseases emerge annually and place enormous burdens on many countries. Additionally, using humans to handle pandemic situations increases the chances of disease spreading. Therefore, various technologies that do not directly involve humans should be employed to handle pandemic situations. The Internet of Drones (IoDT), artificial intelligence (AI), and blockchain are emerging technologies that have revolutionized the modern world. This paper presents a blockchain-based AIempowered pandemic situation supervision scheme in which a swarm of drones embedded with AI is engaged to autonomously monitor pandemic outbreaks, thereby keeping human involvement as low as possible. A use case based on a recent pandemic (i.e., COVID-19) is discussed. Two types of drone swarms are used to handle multiple tasks (e.g., checking face masks and imposing lockdowns). A lightweight blockchain is considered to handle situations in remote areas with poor network connectivity. Additionally, a two-phase lightweight security mechanism is adopted to validate the entities in the proposed scheme. A proof of concept is established using an experimental environment setup and dataset training. An analysis of the experimental results demonstrates the feasibility of the proposed scheme.
The ImageNet Large Scale Visual Recognition Challenge is a benchmark in
object category classification and detection on hundreds of object categories
and millions of images. The challenge has been run annually from 2010 to
present, attracting participation from more than fifty institutions.
This paper describes the creation of this benchmark dataset and the advances
in object recognition that have been possible as a result. We discuss the
challenges of collecting large-scale ground truth annotation, highlight key
breakthroughs in categorical object recognition, provide detailed a analysis of
the current state of the field of large-scale image classification and object
detection, and compare the state-of-the-art computer vision accuracy with human
accuracy. We conclude with lessons learned in the five years of the challenge,
and propose future directions and improvements.
Artificial intelligence of things (AIoT) has brought new promises of efficiency in our daily lives by integrating AI with the IoT. However, owing to limited resources (e.g., computational power), it is difficult to implement modern technology (e.g., AI) and improve its performance (i.e., the IoT). Moreover, cyberthreats and privacy challenges can hinder the success of the IoT. This situation is aggravated by network scarcity (i.e., limited network connectivity). This article presents a digital twin-based data aggregation scheme in which data are collected using federated learning by operating a drone and stored securely in the blockchain. Before data sharing, differential privacy is realized to enhance privacy. A multirole training scheme is proposed, along with a duplex model verification architecture using a Hampel filter and performance check. To validate the specifications, an authentication scheme was implemented by combining a cuckoo filter and timeframe check. A case study to construct an experimental environment using real hardware is discussed. Different experiments were conducted in this environment and the feasibility of the proposed scheme was validated from the outcomes.
Multi-access Edge Computing (MEC) has emerged as a new distributed computing paradigm for its ability to offer low-latency services to users. Suffering from constrained computational resources because of their limited physical sizes, edge servers usually cannot handle all the incoming compute tasks on time when they operate independently. Thus, they need to cooperate by peer-offloading. Incentive and trust are the two major challenges towards to cooperative computing among edge servers operating in a distrusted environment. Another specific challenge in the MEC environment is to facilitate incentive and trust in a decentralized manner. This paper proposes CoopEdge+, a novel blockchain-based decentralized platform, to drive and support cooperative multi-access edge computing to tackle these challenges in a systematic manner. On CoopEdge+, an edge server can publish a compute task for other edge servers to contend for. A winner is selected from candidate edge servers as the task executor based on their reputation to perform the compute task. After that, CoopEdge+ employs a random leader election scheme to elect a task recorder without revealing its leadership until its consensus epoch. The task recorder will coordinate a consensus among edge servers to record the task executor's performance on blockchain. We implement CoopEdge+ based on Hyperledger fabric and evaluate it experimentally against a baseline implementation and three state-of-the-art implementations in a simulated MEC environment. The results validate the usefulness of CoopEdge+ and demonstrate its performance.
The connectivity-aware path design is crucial in the effective deployment of autonomous unmanned aerial vehicles (UAVs). Recently, reinforcement learning (RL) algorithms have become the popular approach to solving this type of complex problem, but RL algorithms suffer slow convergence. In this article, we propose a transfer learning (TL) approach, where we use a teacher policy previously trained in an old domain to boost the path learning of the agent in the new domain. As the exploration processes and the training continue, the agent refines the path design in the new domain based on the subsequent interactions with the environment. We evaluate our approach considering an old domain at sub-6 GHz and a new domain at millimeter-wave (mmWave). The teacher path policy, previously trained at the sub-6 GHz path, is the solution to a connectivity-aware path problem that we formulate as a constrained Markov decision process (CMDP). We employ a Lyapunov-based model-free deep
Q
-network (DQN) to solve the path design at sub-6 GHz that guarantees connectivity constraint satisfaction. We empirically demonstrate the effectiveness of our approach for different urban environment scenarios. The results demonstrate that our proposed approach is capable of reducing the training time considerably at mmWave.
New diseases (e.g., monkeypox) are showing up and taking the form of a pandemic within a short time. Early detection can assist in reducing the spread. However, because of privacy-sensitive data, users do not share it continually. Thus, it becomes challenging to employ modern technologies (e.g., deep learning). Moreover, cyber threats encircle both communication and data. This paper introduces a blockchain-based data acquisition scheme during the pandemic in which federated learning (FL) is employed to assemble privacy-sensitive data as a form of the trained model instead of raw data. A secure training scheme is designed to mitigate cyber threats (e.g., man-in-the-middle-attack). An experimental environment is formulated based on a recent pandemic (i.e., monkeypox) to illustrate the feasibility of the proposed scheme.
Federated Learning (FL) has recently become an effective approach for cyberattack detection systems, especially in Internet-of-Things (IoT) networks. By distributing the learning process across IoT gateways, FL can improve learning efficiency, reduce communication overheads and enhance privacy for cyberattack detection systems. However, one of the biggest challenges for deploying FL in IoT networks is the unavailability of labeled data and dissimilarity of data features for training. In this paper, we propose a novel collaborative learning framework that leverages Transfer Learning (TL) to overcome these challenges. Particularly, we develop a novel collaborative learning approach that enables a target network with unlabeled data to effectively and quickly learn “knowledge” from a source network that possesses abundant labeled data. It is important that the state-of-the-art studies require the participated datasets of networks to have the same features, thus limiting the efficiency, flexibility as well as scalability of intrusion detection systems. However, our proposed framework can address these problems by exchanging the learning “knowledge” among various deep learning models, even when their datasets have different features. Extensive experiments on recent real-world cybersecurity datasets show that the proposed framework can improve more than 40% as compared to the state-of-the-art deep learning based approaches.
This letter presents a federated learning-based data-accumulation scheme that combines drones and blockchain for remote regions where Internet of Things devices face network scarcity and potential cyber threats. The scheme contains a two-phase authentication mechanism in which requests are first validated using a cuckoo filter, followed by a timestamp nonce. Secure accumulation is achieved by validating models using a Hampel filter and loss checks. To increase the privacy of the model, differential privacy is employed before sharing. Finally, the model is stored in the blockchain after consent is obtained from mining nodes. Experiments are performed in a proper environment, and the results confirm the feasibility of the proposed scheme.
In the modern era, the internet of vehicles (IoV) is being utilized in commercial applications and extensively explored in research. However, internal fault in IoV can cause accidents on the road. Moreover, privacy concerns can hamper the internal data sharing to build a model to detect the anomaly. Federated learning (FL) and blockchain are emerging technologies that can assist in mitigating these challenges. FL-based anomaly detection is introduced to prevent road accidents with the help of blockchain. An environment is built to conduct experiments to prove the feasibility of the proposed scheme. The performance analysis demonstrates that our presented scheme outperforms the traditional scheme while having privacy concerns.
Internet of Things (IoT) established the foundation of the smart world via its diverse functionalities (e.g., remote data acquisition, automated task execution). This trend continues to military applications too by introducing internet of military things (IoMT). Due to being at infancy level, new technologies require to move forward and assist IoMT to gain maturity. Mixed reality (MR) can be a potential technology which fuses the virtual and actual world. MR can improve the quality of service (QoS) in terms of inventory management, remote mission handling, battlefield assistance, etc. However, data among these applications is surrounded by cyber threats (e.g., illegal data modification, unauthorised data access). Blockchain is another promising technology which brings security in the distributed world. A content sharing scheme for MR application is proposed on the top of blockchain to bring security in the multi-user environment in IoMT. A smart contract is employed to manage security in accessing data by different users. Moreover, an experimental environment is set to observe the performance of the proposed scheme. The analysis manifests that proposed scheme maintains security without affecting the regular performance.
Modern mobile devices have access to a wealth of data suitable for learning models, which in turn can greatly improve the user experience on the device. For example, language models can improve speech recognition and text entry, and image models can automatically select good photos. However, this rich data is often privacy sensitive, large in quantity, or both, which may preclude logging to the data-center and training there using conventional approaches. We advocate an alternative that leaves the training data distributed on the mobile devices, and learns a shared model by aggregating locally-computed updates. We term this decentralized approach Federated Learning. We present a practical method for the federated learning of deep networks that proves robust to the unbalanced and non-IID data distributions that naturally arise. This method allows high-quality models to be trained in relatively few rounds of communication, the principal constraint for federated learning. The key insight is that despite the non-convex loss functions we optimize, parameter averaging over updates from multiple clients produces surprisingly good results, for example decreasing the communication needed to train an LSTM language model by two orders of magnitude.
Recent reports suggest that a generic supervised deep CNN model trained on a
large-scale dataset reduces, but does not remove, dataset bias on a standard
benchmark. Fine-tuning deep models in a new domain can require a significant
amount of data, which for many applications is simply not available. We propose
a new CNN architecture which introduces an adaptation layer and an additional
domain confusion loss, to learn a representation that is both semantically
meaningful and domain invariant. We additionally show that a domain confusion
metric can be used for model selection to determine the dimension of an
adaptation layer and the best position for the layer in the CNN architecture.
Our proposed adaptation method offers empirical performance which exceeds
previously published results on a standard benchmark visual domain adaptation
task.
An improved analysis of stochastic gradient descent with momentum
- Y Liu
- Y Gao
- W Yin
Y. Liu, Y. Gao, and W. Yin, "An improved analysis of stochastic gradient
descent with momentum," in Proceedings of the 34th International
Conference on Neural Information Processing Systems, ser. NIPS'20.
Red Hook, NY, USA: Curran Associates Inc., 2020.
An improved analysis of stochastic gradient descent with momentum
- Liu