Jian Yang, Xun Liu, Xiaofeng Jiang, Yaohui Zhang, Shuangwu Chen, Huasen He
©SHUTTERSTOCK.COM/CHESKY
The unmanned aerial vehicle (UAV) swarm is widely applied in many fields, such as disaster relief and geological exploration. However, it is highly vulnerable when facing an attack from internal malicious UAVs due to the underlying disconnected, intermittent, and limited (DIL) network environment. Recently, building trusted networks has been considered as a strategy to address security issues. Nevertheless, deploying a trusted network in a swarm faces many challenges imposed by the dynamic topology and limited resources of DIL networks. In order to overcome these challenges, we developed a blockchain-based UAV trusted self-organizing networking mechanism (BC-UTSON) to enhance internal security. Three key technologies are proposed for assisting BC-UTSON in achieving trustworthiness. UAV practical Byzantine fault tolerance (U-PBFT) is proposed to guarantee the lightweight consensus and real-time full-node trustworthiness evaluation under the hierarchical self-organizing networking structure of a UAV swarm. A blockchain-based multiweighted subjective logic (BMWSL) scheme has been designed to identify malicious UAV nodes based on the trustworthiness evaluation, which also assists the trusted path quality-aware dynamic routing (TPDR) mechanism to prevent data from routing through malicious nodes. The experimental results show that the proposed BC-UTSON performs much better than nonblockchain-based mechanisms for the UAV swarm.
The UAV swarm is a collaborative system composed of a large number of multifunction UAVs with a self-organizing network and swarm intelligence. Networked UAV swarm enables multiple UAVs in a cooperative manner to perform complex tasks, such as secure data collection [1] and collaborate federated learning [2]. Thus, the UAV swarm network is envisioned to be an essential part of tactical wireless communications. Nevertheless, security is an important factor in preventing extensive equipment of UAV swarms. The swarm network is highly vulnerable when facing attacks from the internal malicious UAVs due to the underlying DIL network environment. Attacks can easily launch routing interruption attacks in swarm networks through manipulated UAVs, such as black-hole attack and gray-hole attack, which drop received data packets silently. In self-organizing networks, a few black-hole UAVs can drastically reduce the packet delivery rate of the whole network.
Hence, building trusted networks is highly attractive as an effective method to achieve network security. In the trusted network, the trustworthiness is determined by the interactions of swarm nodes. Only the trusted nodes are allowed to participate in the network decisions and to provide services for other swarm nodes, while the malicious nodes are isolated from the network [3].
Recently, for the network without a centralized trustworthiness manager, blockchain has been introduced to build decentralized trusted systems, which can provide verifiable and traceable interaction records. Our main objective is to build a trusted UAV swarm self-organizing network by deeply combining trustworthiness management and blockchain. However, the underlying DIL network environment is largely different from traditional fixed-ground networks, and there are still three key challenges that need to be addressed.
The first challenge is to maintain the performance of the blockchain under a traffic concentration network structure. The UAV swarm network is usually divided into communication clusters according to a hierarchical structure [4], where the data between different clusters are forwarded by the cluster head nodes (CHNs), and CHNs are thus the performance bottleneck of swarm networks. The peer-to-peer (P2P)-based broadcast scheme of the blockchain consumes too much intercluster bandwidth, which leads to network congestion at CHNs. The second challenge is to achieve precise trustworthiness evaluation under malicious node disturbance. In several scenarios, such as package delivery and object detection, the UAVs manipulated by malicious parties can be used for espionage or damage the network reliability [5]. The blockchain-based trust management mechanism is regarded as an effective, lightweight method to protect the internal security of systems. However, most trust management mechanisms only use the blockchain as a trusted ledger to record the node opinions. The sensitivity of trustworthiness evaluation is low and easily disturbed by misleading information. The third challenge is to implement secure data routing under dynamic network topology. The operation of blockchains depends on the network connectivity, and the disturbance with data transmission can lead to performance degradation of blockchains [6]. It is necessary to design a routing protocol to ensure the security and low packet loss rate of the transmission path.
To overcome the above challenges, we developed a trusted self-organizing networking mechanism with built-in blockchain for UAV swarm, namely BC-UTSON, which consists of the U-PBFT, BMWSL, and TPDR. U-PBFT is a two-layer BFT consensus mechanism corresponding to the hierarchical structure of the swarm network and has high scalability, while retaining the instant finality of single-layer BFT. BMWSL is used to precisely evaluate the full-node trustworthiness when the malicious nodes are not accurately identified. The UAV behaviors are divided into verifiable behaviors and unverifiable behaviors to improve the accuracy and sensitivity of trustworthiness evaluations. TPDR is a distributed self-organizing network routing protocol that ensures the connectivity of the underlying network by selecting transmission paths with low packet loss rates and high security. Our contributions are summarized as follows:
The rest of this article is organized as follows. The subsequent “Related Work” section is followed by the “Key Technologies for BC-UTSON” section. Next, the “Performance Evaluations” section conducts experiments to evaluate our proposed mechanism, followed by the conclusion in the last section.
Blockchain is a data structure composed of data blocks in chronological order, and it is also a ledger that is decentralized, traceable, and tamper-proof [7]. With the improvement of hardware performance and the miniaturization of airborne devices, some lightweight blockchains are deployed in UAVs to ensure the safety of swarm systems under the limited bandwidth with the ground station. Aujla et al. [8] use the blockchain to ensure secure communication between UAVs to defend against spoofing attacks, GPS attacks, and other attacks. Tang et al. [1] utilize the blockchain to assist Internet of Things devices for secure data collection. These works ignore the impact of network structure on blockchain performance, which is the concern of our article. The proposed U-PBFT optimizes data communication based on the swarm network structure and thus can be deployed in large-scale swarm networks. In addition, blockchain performance is determined by the communication quality of the underlying network. The transmission path should avoid malicious nodes and have a low packet loss rate to avoid frequent retransmissions. Commonly used ad hoc network routing protocols, such as optimized link state routing protocol (OLSR) [9] and ad hoc on-demand distance vector protocol (AODV) [10], pay more attention to communication delays. In contrast, our proposed TPDR routing protocol focuses on link quality and security, and thus can provide higher network throughput for blockchains.
In the trust management mechanism, the blockchain serves as a trusted ledger to record and aggregate node opinions. Barka et al. [11] used blockchain to record the node opinions uploaded by the miners and provide data for trust evaluation. Zhang et al. [12] utilized the blockchain to evaluate the trustworthiness of road side units (RSUs) and detect malicious RSUs in the Internet of Vehicles. For enhancing the accuracy and sensitivity under low-frequency interactions, Li et al. [13] used trusted RSUs to actively send probe packets, which increased the interaction frequency and thus improved the trustworthiness evaluation result. However, these works only used the blockchain as a trusted ledger to record opinions, and the evaluation results may still be disturbed by misleading information. We use the blockchain consensus to replace the trustworthiness exchange between nodes to eliminate misleading information disturbance to achieve higher evaluation accuracy and sensitivity.
The BC-UTSON mechanism contains three key technologies, i.e., U-PBFT for improving the consensus efficiency and reducing the network congestion, BMWSL for identifying malicious swarm nodes, and TPDR for protecting the network connectivity.
The UAV swarm network is usually deployed in a hierarchical structure to achieve scalability and low communication delay. As shown in Figure 1, the UAV swarm is divided into a series of communication clusters based on their locations. In each cluster, a UAV with sufficient resources and high security is selected as the CHN for coordinating intercluster routing and data transmission. However, such a traffic-centralized two-layer communication structure cannot meet the single-layer P2P broadcast requirements of conventional blockchains, where frequent intercluster broadcasts exacerbate congestion at CHNs. The BFT consensus principle can ensure consistency: i.e., forks are almost impossible and instant finality is granted, which means that the data in the new block are reliable and can be used for time-sensitive decisions immediately. Based on PBFT, we propose a novel consensus mechanism named U-PBFT to provide quick blockchain confirmation with low intercluster bandwidth consumption. The specific process is shown in Figure 1, which is described as follows.
Figure 1 The hierarchical consensus process of the UAV swarm network. CLN, consensus leader node.
The transactions in the U-PBFT mechanism are divided into clusters, interaction, and ordinary transactions according to the transaction_type field. The cluster transactions are submitted to support the mobility management of blockchain nodes when UAVs move to other communication clusters. After each new block is generated, the UAVs use the cluster transactions to update the communication cluster node lists. The interaction transactions contain verifiable interactions, such as access control and anomalous voting, which are utilized for trustworthiness evaluation. The ordinary transactions are the task and resource information released according to the swarm’s purpose.
U-PBFT requires the UAVs to obtain identity information (ID, public key, etc.) authenticated by trusted participants before joining the swarm. Then, the UAVs pack the identity information, the trusted participants’ signatures, and the joined communication cluster ID into a transaction and broadcast it to the swarm. After approval by consensus, the UAVs are allowed to join the swarm network.
The cluster transaction is also used for exiting of UAVs, which means writing the target_cluster field as EXITING. Smart contracts are used for the mobility management of UAVs, initialized by the ground station. In addition to the join and exit functions of the UAVs, the smart contract also includes system parameter setting, UAV registration, UAV leaving, UAV joining, UAV revocation, UAV migration, swarm task decision making, and UAV behavior auditing, etc.
Similar to the CHNs, U-PBFT selects a trustworthy node as the consensus leader node (CLN) in each cluster for coordinating block verification with the consensus process. Since serving as CHN and CLN will introduce additional computing and communication overhead to the UAV, we stipulate that CHN and CLN should be served by different UAVs to reduce node energy consumption and improve system robustness. Besides, other UAVs need to monitor the status of the leader UAVs to ensure normal consensus and communication. The consensus process is a two-layer PBFT model: intracluster consensus led by CLNs and intercluster consensus led by a CLN leader. Each phase includes three stages: preprepare, prepare, and commit, as shown in Figure 1. In the swarm network, UAVs upload the information as transactions for verification and announcement. After collecting the transactions, the leader broadcasts preprepare messages to other nodes. In the prepare and commit stages, all nodes broadcast messages to consensus the validity of transactions. In contrast to the conventional single-layer PBFT consensus, U-PBFT uses intercluster links only in the intercluster consensus step. Since the CLNs are only in a small fraction of all consensus nodes, U-PBFT can significantly reduce the load on intercluster links and improve the consensus efficiency of the whole swarm network. Relatively, the hierarchical structure makes the security of U-PBFT lower than that of PBFT. In the worst case, U-PBFT can tolerate up to one-sixth of UAV failures and intrusions. However, we can improve the security of the system in combination with the trust management mechanism.
The blockchain trades data redundancy for system security, and putting excessive data on the chain will degrade system performance. Thus, we only use the blockchain to record data for system decisions and data digital for evidence. The increase in block data also needs to be considered. For short-term missions, the block data will not put pressure on data storage. For long-term missions, the data storage problem can be solved by periodic blockchain clipping. Returning UAVs can transfer old block data to the ground storage facility, retaining only the recent block data for decision making. Since instant finality is granted in BFT, the consensus is still reliable. To further reduce intercluster bandwidth consumption of broadcasting transactions, the P2P-based broadcast is replaced by layered broadcast at the network layer, as shown in Figure 1, where messages are transmitted to physical neighbors instead of P2P neighbors, and the intercluster broadcast is only performed at the CHN layer. This efficient broadcast scheme can greatly reduce bandwidth consumption, which allows all UAVs to hold block data, thereby improving the robustness of the swarm system.
The subjective logic model is a popular distributed trustworthiness evaluation model with the advantages of low energy consumption and easy deployment [14]. Based on this, we propose the BMWSL scheme, which utilizes blockchain technology to detect malicious UAVs accurately and sensitively.
In the BMWSL scheme, we divide the UAV behaviors into two parts: verifiable behaviors and unverifiable behaviors. The verifiable behaviors include consensus votes, content access requests, etc. When such anomalous behaviors are caught, they will be uploaded to the blockchain for review by the smart contract, which audits the behaviors according to established rules stored in the blockchain. Then, the audit results are written into blocks as transactions and broadcast to the whole network. Once the behavior is audited as malicious, all UAVs will downgrade the trustworthiness of the related UAV, even if there is no interaction between them. The trustworthiness computed from verifiable behaviors is defined as ${r}_{v} = {{\omega}_{s}{n}_{s}}\slash{{\omega}_{s}{n}_{s} + {\omega}_{f}{n}_{f}}$, where ${n}_{s},{n}_{f}$ are the number of positive and negative behaviors, respectively, and ${\omega}_{s},{\omega}_{f}$ are the corresponding weights, which are set during system initialization. Considering that negative behaviors are usually less frequent and more influential than positive behaviors, we suggest setting the behavior weights ${\omega}_{s}\,{<}\,{\omega}_{f}$. Compared with the trust evaluation mechanism that only uses the blockchain to record and aggregate trustworthiness, BMWSL utilizes smart contracts to achieve credible trustworthiness evaluation of ${r}_{v}$, eliminating evaluation errors caused by distrust of indirect opinions. Besides, since the parameters for calculating ${r}_{v}$ are stored in the blockchain, the malicious nodes cannot disturb the evaluation by spreading misleading trustworthiness.
The unverifiable behaviors cannot be uploaded to the blockchain since their authenticity cannot be verified. Therefore, UAVs need indirect opinions to calculate the trustworthiness of the target, which is defined as ${r}_{nv} = {\alpha}{r}_{\text{dir}} + {(}{1}{-}{\alpha}{)}{r}_{\text{indir}}$, where ${\alpha}$ is the confidence factor of direct opinion. UAVs evaluate the trustworthiness of target UAVs by exchanging opinion packets. The evaluation of ${r}_{\text{dir}}$ is similar to ${r}_{v}$ and ${r}_{\text{indir}}$ is a weighted sum of the opinions of other UAVs according to the trustworthiness. However, only exchanging opinions cannot provide sufficient evaluation sensitivity and is susceptible to misleading opinions. For higher accuracy and sensitivity, we introduce the active detection mechanism to further improve the detection accuracy. As shown in Figure 2, when the interaction frequency is low or the variance of indirect opinions is large, the UAV will actively test the neighbor’s security using probe packets to improve the evaluation accuracy. The probe packet is disguised as a data packet, while the probe flag is hidden in the symmetric key encrypted data segment. If the destination UAV receives the probe packet, it will feed it back to the source UAV within a specified time. Furthermore, we limit the TTL of indirect opinion packets to two to prevent the collusion of scattered malicious nodes. When relaying the opinion packets, UAVs only forward the packets from a neighbor to filter redundant information. Since swarm networks usually adopt ad hoc routing, the neighbors’ security directly determines the security of data transmission. The above improvements can prevent the collusion of malicious nodes and accurately and quickly evaluate neighbors’ security to resist black-hole attacks.
Figure 2 Examples of the probe and opinion packets.
Considering that ${r}_{v}$ can be regarded as the “probability” that a node is normal in terms of verifiable behavior, and ${r}_{nv}$ can be regarded as the “probability” that a node is normal in terms of unverifiable behavior, we define the final trustworthiness r as ${r} = {r}_{v}\,{\times}\,{r}_{nv}$, which represents the probability that a node behaves normally. Compared with the weighted sum-based evaluation criterion, the product-based evaluation criterion has higher sensitivity in identifying malicious nodes.
Due to the high-mobility of UAV nodes, UAV swarm networks usually use self-organizing routing protocols for data transmission to enhance network reliability. However, the high mobility of UAV nodes not only makes the network topology change rapidly but also causes the network security environment to change frequently. This means that malicious nodes may appear on the routing path during data transmission. If the routing path is not switched in time, the data will be at risk of being lost or tampered with. Unfortunately, current wireless self-organizing routing protocols only consider link states and transmission hops, ignoring the security of the transmission path. To address the challenge induced by the frequent change of the security environment, we propose the TPDR mechanism with the aid of the trustworthiness evaluation scheme, considering both the path quality and the security of the relay nodes. TPDR is a distributed routing protocol in which each node determines only the next hop of data forwarding rather than the entire transmission path. This allows the secure transmission path to dynamically change in time to avoid malicious nodes while keeping the low routing overhead.
In TPDR, the UAV node only determines the optimal next hop based on the STQ rather than the entire transmission path. The single-hop STQ is defined as the product of the transmission quality (TQ) of the path and the trustworthiness of the endpoint node. Here, TQ represents the transmission success rate of the path. According to BMWSL, the nodes with low trustworthiness are likely to be malicious nodes. Thus, a low STQ value means that the transmission success rate of the path is low, and there may be malicious nodes in the path. The calculation of the STQ is achieved by broadcasting the routing learning message (RLM). In the UAV swarm network, each node periodically broadcasts the RLM to neighbor nodes and calculates the STQ by counting the received RLM.
Figure 3 illustrates the calculation of the single-hop STQ. A pair of sliding windows, namely the reception quality (RQ) window, and the loop quality (LQ) window, are utilized for calculating the single-hop STQ. The RQ window counts the number of RLMs that are generated by the local node and rebroadcasted. The RQ window slides along the maximum index of the received RLM to guarantee the timeliness of RQ. The LQ window counts the number of RLMs generated by the local node and rebroadcasted back to the local node by the neighbor node. The LQ value represents the successful transmission rate of the loop spanning from the local node to its neighbor and then back to the local node. The LQ window slides periodically with RLM generation, since the LQ messages are generated by the local node. When node A knows the RQ and LQ value of the link from node A to node B, it can straightforwardly calculate the single-hop TQ by the ratio of LQ to RQ. Furthermore, since LQ and RQ measurements are independent of each other, there may be cases that LQ is higher than RQ. Here we set TQ to 1 for this case. After calculating the TQ, node A looks up the behavior records of node B in the block and calculates the trustworthiness of B according to BMWSL. By combining TQ and trustworthiness, the single-hop STQ can be calculated, which comprehensively characterizes the probability of data being safely transmitted from A to B. Then the multihop STQ can be obtained by multiplying single-hop STQs corresponding to Figure 3.
Figure 3 Calculation of the STQ and secure transmission process. LQ, loop quality; TQ, transmission quality; RQ, reception quality.
After determining the optimal transmission path through the above routing scheme, the network topology changing and nodes being attacked may cause the routing path to suddenly become unsecured. In this case, routing data using this unsecured path may cause data to be maliciously dropped and manipulated. Motivated by this, by monitoring the trustworthiness of the next hop, the data can be forwarded only if the trustworthiness is higher than a given threshold value. Otherwise, the data will be cached and rerouted. Figure 3 illustrates an example, where node A is transmitting data to node E along the initial path A-B-C-D-E. It first transmits the data to B, and then node B selects C as the next hop. When node C prepares to transmit the data to node D, it observes that the trustworthiness of node D suddenly drops. Note that at this time, node B is unaware of the sudden decrease, so it will continue to transmit the data to node C. Then node C caches the data and reroutes, and finds that node B is the optimal next-hop node. Hence node C sends data back to B and informs that this path is not available. Node B reroutes the data and sets node F as the next-hop node. Finally, node A transmits data to E via the new transmission path A-B-F-G-E.
Finally, we analyze the computational complexity of BC-UTSON. Consider a swarm system with N UAVs divided into M clusters, including ${N}_{1},{\ldots}\,{N}_{M}$ UAVs, respectively. If the handling of a message is regarded as an operation, the computational complexity of U-PBFT is ${O}{(}{N}_{1}^{2} + {\cdots} + {N}_{M}^{2} + {M}^{2}{)}$, the computational complexity of BMWSL is ${O}{(}{N}_{1}^{2} + \cdots + {N}_{M}^{2} + {M}{(}{N} - {M}{)}{)}$, and the computational complexity of TPDR is ${O}{(}{N}_{1}^{3} + {N}_{2}^{3} + \cdots + {N}_{M}^{3} + {M}^{3}{)}$.Importantly, the computational complexity of BMWSL and TPDR considers fully connected topologies, which are almost impossible in real environments. In general topologies, the computational complexity of BMWSL and TPDR is much lower than in fully connected scenarios.
In this section, we carried out experiments to evaluate the performance of the proposed BC-UTSON in the scenario where the swarm network may be disturbed by black-hole attacks. We cooperated with Shanghai Energy Chain Co., Ltd. to develop a blockchain platform that combines blockchain with self-organizing network protocols. As shown in Figure 4(a), the swarm system contains six normal UAV nodes and two malicious UAV nodes. The UAV node includes a DJI M300 RTK UAV, which is equipped with a DJI Manifold-2G onboard computer and a SUNTOR ST9633 wireless communication device. The software of BC-UTSON is deployed on the Manifold-2G.
Figure 4 Trust self-organizing network experiment in real-world environment. (a) The evaluation of the performance of BMWSL. (b) The evaluation of the energy consumption of BC-UTSON. (c) The evaluation of the network throughput of TPDR.
Before starting the experiment, each node runs for a while to maintain trustworthiness at nearly 0.6. In the first 20 s, each node behaves well to obtain positive opinions from other nodes. After this, malicious nodes randomly interrupt the traffic from normal nodes. These normal nodes generate negative behaviors for the malicious nodes when they detect the black-hole attack. Therefore, the trustworthiness of the malicious node gradually decreases after 20 s. Figure 4(a) depicts the data transmission path from node C to node H in the swarm network. Nodes A and F are malicious nodes, where F maliciously drops the forwarded packets, and A publishes misleading trustworthiness evaluation information. The bottom of Figure 4(a) shows the trustworthiness variation of node F from the perspective of node C. For the MWSL [14] and BMWSL, the trustworthiness decreases sharply due to the opinions and records from the other nodes, and the trustworthiness below the secure threshold of the next transmission hop node occurs at the 30th and 39th seconds, respectively. When the trustworthiness of F is less than 0.4, the TPDR switches the transmission path from C-F-H to C-D-G-H to guarantee security. Compared with TSL and MWSL, our scheme can switch paths more quickly because it not only takes into account timeliness and behavior influence but also uses smart contracts to combine conflicting records to eliminate misleading information.
To evaluate the impact of BC-UTSON on the UAV life span, we measured the power of the airborne computer. The experimental results are shown in Figure 4(b). Compared to the flight energy consumption, the energy consumption of BC-UTSON is slight. The average idle power of the airborne computer is 6.73 W, the average power of BC-UTSON software running without data transmission is 7.33 W, and the average power of BC-UTSON software running data transmission is 7.42 W, which is only an increase of 0.69 W compared to idle power (about 10%). In addition, the transmitting power of the wireless communication device is 1–2 W. The experimental UAV contains two 274 Wh batteries, the longest life span is 55 min, and the flight power is estimated to be 598 W, which is far larger than the energy consumed by BC-UTSON. Finally, we compared the network throughput of TPDR with the benchmark ad hoc network routing algorithms OLSR [9] and AODV [10], which can also reflect the routing performance. The experimental results are shown in Figure 4(c). In the beginning, all three routing algorithms choose the same path to transmit data. At the 75th second, the original link is cut off, OLSR and AODV switch to the shortest but unstable path, while TPDR switches to the path with the best path quality and thus has a higher network throughput.
We have demonstrated the performance of TPDR and BMWSL in a real-world environment, but the performance of U-PBFT needs to be tested in a large-scale network. Thus, we used Kubernetes to build a semiphysical simulation platform. The sim-physical simulation platform includes a controller and hundreds of containers. The container contains the same BC-UTSON software as in UAV, performing realistic data transmission and routing in simulation. The controller calculates the network topology and link quality based on the configuration information and sends them to each container to simulate the real environment. We compared our U-PBFT with conventional PBFT [15]. Figure 5(a) shows the average performance of U-PBFT under different topologies. The block confirmation time and transaction throughput of U-PBFT remain stable as the network scale increases, while the block confirmation time of PBFT increases sharply. The reason is reflected in Figure 5(b), U-PBFT can still maintain low queue length at CHNs in a network with hundreds of nodes by clustering and optimizing the broadcast strategy. In contrast, when the number of UAVs exceeds 110, congestion occurs at CHNs since the transmission rate cannot meet the consensus requirements.
Figure 5 Performance evaluations of consensus protocols. (a) The evaluation of the blockchain performance. (b) The evaluation of the blockchain network resource consumption.
In this article, we have proposed a trusted self-organizing networking mechanism called BC-UTSON, which is deployable in a UAV swarm to secure data and decisions. In the BC-UTSON, the U-PBFT improves the consensus efficiency of the blockchain and reduces the network load caused by consensus. The BMWSL scheme evaluates node security and removes malicious nodes from the swarm network. The TPDR based on evaluation results of the BMWSL scheme is presented to guarantee the security of the swarm network. Our future work will focus on the deployment and optimization of BC-UTSON, and we will release the BC-UTSON work to the China Environment for Network Innovations experimental platform after testing.
Xiaofeng Jiang would like to acknowledge the financial support of National Natural Science Foundations of China (Grants 62173315, 62101525, 62021001), Youth Innovation Promotion Association Chinese Academy of Sciences (Grant 2020450), Strategic Priority Research Program of Chinese Academy of Sciences (Grant XDC07020200), and Fundamental Research Funds for the Central Universities, China Environment for Network Innovations. Xiaofeng Jiang is the corresponding author.
Jian Yang (jianyang@ustc.edu.cn) is a professor in the School of Information Science and Technology, University of Science and Technology of China (USTC), Hefei 230000, China. He received his B.S. and Ph.D. degrees from USTC in 2001 and 2006, respectively. His research interests include future network, distributed system design, modelling and optimization, and multimedia over wired and wireless and stochastic optimization. He received the Lu Jia-Xi Young Talent Award from Chinese Academy of Sciences in 2009.
Xun Liu (abliuxun@mail.ustc.edu.cn) is with the Department of Automation, University of Science and Technology of China (USTC), Hefei 230000, China, and the Institute of Artificial Intelligence, Hefei Comprehensive National Science Center, Hefei 230000, China. He received his B.E. degree in information science and technology from USTC in 2019 and is pursuing his Ph.D. degree at USTC. His research interests include Internet of Things and unmanned aerial vehicle networks.
Xiaofeng Jiang (jxf@ustc.edu.cn) is an associate professor in the School of Information Science and Technology, University of Science and Technology of China (USTC), Hefei 230000, China. He received his B.E. and Ph.D. in information science and technology from USTC in 2008 and 2013, respectively. His recent research interests include discrete event dynamic systems, tensor analysis and big data, and future network and cognitive communications.
Yaohui Zhang (zyh_123@mail.ustc.edu.cn) is with the Department of Automation, University of Science and Technology of China (USTC), Hefei 230000, China, and the Institute of Artificial Intelligence, Hefei Comprehensive National Science Center. He received his B.E. degree from USTC in 2021. He is currently pursuing his masters degree at USTC. His research interests include Internet of Things and unmanned aerial vehicle networks.
Shuangwu Chen (chensw@ustc.edu.cn) is with the Department of Automation, University of Science and Technology of China (USTC), Hefei 230000, China, and the Institute of Artificial Intelligence, Hefei Comprehensive National Science Center. He received his B.S. and Ph.D. degrees from USTC in 2011 and 2016, respectively. He is currently a postdoctoral scholar in the Department of Electronic Engineering and Information Science in USTC. His research interests include multimedia communications, future network, and stochastic optimization.
Huasen He (huasenhe@ustc.edu.cn) is an associate research fellow with the School of Information Science and Technology, University of Science and Technology of China (USTC), Hefei 230000, China. He received his B.S. degree in automation from USTC in 2013, and his M.S. degree in signal processing and communications and Ph.D. degree in digital communications from the University of Edinburgh, Edinburgh, U.K., in 2014 and 2018, respectively. His research interests include future networks, network modelling, and optimization.
[1] X. Tang, X. Lan, L. Li, Y. Zhang, and Z. Han, “Incentivizing proof-of-stake blockchain for secured data collection in UAV-assisted IoT: A multi-agent reinforcement learning approach,” IEEE J. Sel. Areas Commun., vol. 40, no. 12, pp. 3470–3484, Dec. 2022, doi: 10.1109/JSAC.2022.3213360.
[2] S. Dahmane, M. B. Yagoubi, B. Brik, C. A. Kerrache, C. T. Calafate, and P. Lorenz, “Multi-constrained and edge-enabled selection of UAV participants in federated learning process,” Electronics, vol. 11, no. 14, Jul. 2022, Art. no. 2119, doi: 10.3390/electronics11142119.
[3] Z. Yang, K. Yang, L. Lei, K. Zheng, and V. C. Leung, “Blockchain-based decentralized trust management in vehicular networks,” IEEE Internet Things J., vol. 6, no. 2, pp. 1495–1505, Apr. 2019, doi: 10.1109/JIOT.2018.2836144.
[4] L. Gupta, R. Jain, and G. Vaszkun, “Survey of important issues in UAV communication networks,” IEEE Commun. Surveys Tuts., vol. 18, no. 2, pp. 1123–1152, 2nd Quart. 2016, doi: 10.1109/COMST.2015.2495297.
[5] B. Bera, A. K. Das, and A. K. Sutrala, “Private blockchain-based access control mechanism for unauthorized UAV detection and mitigation in internet of drones environment,” Comput. Commun., vol. 166, pp. 91–109, Jan. 2021. [Online] . Available: https://www.sciencedirect.com/science/article/pii/S0140366420320119, doi: 10.1016/j.comcom.2020.12.005.
[6] R. J. Cai, X. J. Li, and P. H. J. Chong, “An evolutionary self-cooperative trust scheme against routing disruptions in MANETs,” IEEE Trans. Mobile Comput., vol. 18, no. 1, pp. 42–55, Jan. 2019, doi: 10.1109/TMC.2018.2828814.
[7] Y. Tan, J. Liu, and N. Kato, “Blockchain-based lightweight authentication for resilient UAV communications: Architecture, scheme, and future directions,” IEEE Wireless Commun., vol. 29, no. 3, pp. 24–31, Jun. 2022, doi: 10.1109/MWC.002.2100594.
[8] G. S. Aujla, S. Vashisht, S. Garg, N. Kumar, and G. Kaddoum, “Leveraging blockchain for secure drone-to-everything communications,” IEEE Commun. Standards Mag., vol. 5, no. 4, pp. 80–87, Dec. 2021, doi: 10.1109/MCOMSTD.0001.2100012.
[9] T. Clausen and P. Jacquet, “Optimized link state routing protocol (OLSR),” Internet Engineering Task Force, Fremont, CA, USA, RFC3626, 2003.
[10] C. Perkins, E. Belding-Royer, and S. Das, “Ad Hoc on-demand distance vector (AODV) routing,” Internet Engineering Task Force, Fremont, CA, USA, RFC3561, 2003.
[11] E. Barka, C. A. Kerrache, H. Benkraouda, K. Shuaib, F. Ahmad, and F. Kurugollu, “Towards a trusted unmanned aerial system using blockchain for the protection of critical infrastructure,” Trans. Emerg. Telecommun. Technol., vol. 33, no. 8, 2022, Art. no. e3706, doi: 10.1002/ett.3706.
[12] H. Zhang, J. Liu, H. Zhao, P. Wang, and N. Kato, “Blockchain-based trust management for Internet of Vehicles,” IEEE Trans. Emerg. Topics Comput., vol. 9, no. 3, pp. 1397–1409, Jul./Sep. 2021, doi: 10.1109/TETC.2020.3033532.
[13] F. Li, Z. Guo, C. Zhang, W. Li, and Y. Wang, “ATM: An active-detection trust mechanism for vanets based on blockchain,” IEEE Trans. Veh. Technol., vol. 70, no. 5, pp. 4011–4021, May 2021, doi: 10.1109/TVT.2021.3050007.
[14] J. Kang, Z. Xiong, D. Niyato, D. Ye, D. I. Kim, and J. Zhao, “Toward secure blockchain-enabled internet of vehicles: Optimizing consensus management using reputation and contract theory,” IEEE Trans. Veh. Technol., vol. 68, no. 3, pp. 2906–2920, Mar. 2019, doi: 10.1109/TVT.2019.2894944.
[15] M. Castro et al., “Practical byzantine fault tolerance,” in Proc. 3rd Symp. Oper. Syst. Des. Implementation (OsDI), 1999, vol. 99, pp. 173–186.
Digital Object Identifier 10.1109/MVT.2023.3242834