Federated learning (FL) in Unmanned Aerial Vehicle (UAV) swarms faces the challenges of data heterogeneity and resource constraints, limiting its large-scale deployment. Existing solutions attempt to leverage Clustered Federated Learning (CFL) to mitigate data heterogeneity by grouping clients based on data similarity. However, these data-driven approaches concentrate on application-layer features without comprehensive consideration of status in other layers, leading to unstable clustering and suboptimal routing. In this paper, we propose X-CFL, a novel cross-layer framework that co-optimizes clustering and routing by integrating application-layer data features with cross-layer node status. Specifically, X-CFL introduces a joint clustering mechanism that groups UAVs based on data similarity to minimize model discrepancy, while concurrently leveraging real-time physical conditions (e.g., location and energy) to enhance cluster robustness. Subsequently, a two-stage routing protocol is employed to establish reliable intra-cluster communication and enable efficient global aggregation among cluster heads. Extensive simulation results demonstrate that X-CFL significantly improves training throughput, network lifetime, and model training performance compared to state-of-the-art baselines.

Practical applications in mobile ad-hoc networks (MANETs) require the support of diverse network services, e.g., host-centric, content-centric, and location-centric routing and forwarding services. However, existing solutions are typically designed over a single network service rather than integrated ones. To embed diverse network services into MANETs, the major challenge is enabling interoperability among various network-layer (L3) protocols without suffering complexity and scalability issues. In this article, we propose OpenL3, a programmable L3 approach to support the coexistence of diverse network services in MANETs. Specifically, OpenL3 first abstracts key attributes from network entities, such as content, locations, or groups of devices. These attributes are embedded into a network address, named multidimensional identifier (MID), to control the routing and forwarding processes. Then, a distributed MID mapping system is established to facilitate efficient MID registration and query. Based on the MID, a programmable routing and forwarding scheme is proposed, which incorporates a lightweight packet processing design using a P4 programmable data plane to enable interoperability among various L3 protocols. A cluster of SDN-based control plane devices collaboratively distribute flow rules to manage data plane behavior. Furthermore, a prototype system is built to implement and evaluate the proposed solutions. Experimental results show that OpenL3 outperforms the existing solutions in terms of end-to-end latency and network throughput while being deployable in MANETs without modifications to network protocols or sockets.

Geo-Distributed Machine Learning (GDML) aims to train large-scale machine learning models across geographically dispersed datacenters. However, the performance of GDML systems is constrained by the limited Wide Area Network (WAN) bandwidth and the presence of the straggler problem. Existing GDML designs often show contradictory effects in addressing these challenges, while in-network computing attempts are typically restricted to single datacenter environments rather than the more complex GDML scenarios. To overcome these limitations, this paper proposes L3DML to facilitate GDML using the P4-based Software-defined Network (SDN). Our approach incorporates three key innovations. Firstly, we introduce a novel network addressing scheme that enables location-specific in-network gradient aggregation for GDML, eliminating the need for parameter servers. Secondly, we utilize the P4 data plane to integrate lossless gradient transmission within switches. Thirdly, we address the straggler problem by employing a unique Deep Reinforcement Learning (DRL) model set and a corresponding rate synchronization routing approach. L3DML is implemented on a prototype system consisting of several Intel Tofino switches and the Spirent network emulator. Experimental results indicate that L3DML outperforms existing solutions in terms of goodput, model accuracy, and training speed gain for large-scale GDML.

Recent research has demonstrated that the Computing Aware Network (CAN) effectively balances network resources and computing requirements. It achieves this by intelligently directing traffic to suitable computing resources, taking into account both routing metrics and computing resource metrics. However, attackers may publish false advertisements regarding computing resources for various purposes, which cause traffic hijack and a decrease in network computing performance. This poses significant challenges in evaluating and trusted publishing of computing resources. To address this issue, this paper proposes a blockchain-based Computing Resource Authenticity Evaluation solution (CRAE), which utilizes blockchain and game theory to enhance the usage of computing resources and availability of CAN. First, a bilateral evaluation mechanism based on game theory is designed to suppress the impact of false advertisements of computing resources on the overall efficiency of the network. Second, based on the immutability of blockchain technology, a public ledger has been designed for trusted publishing of computing resources. Finally, the simulation results indicate that compared to existing security mechanisms, CRAE can effectively suppress malicious behavior in the network and reduce the waste of computing resources resulting from the introduction of security mechanisms.

Geocast is a one-to-many communication paradigm that enables the transmission of data packets to a designated area rather than an IP address. The most common geocast solutions rely on the application-layer Geolocation-to-IP database. But these IP-based approaches cannot cope with the challenges of flexibility and mobility in a granularity-customizable geocast scenario. While some non-IP network-layer (L3) attempts have resulted in low addressing accuracy and poor routing scalability. Besides, the clean-slate design is incompatible with the existing network. To address these issues, this paper proposes an innovative network-layer geographic addressing scheme that leverages P4-based Software Defined Networks (SDN) to enable flexible geocast with high accuracy. Based on the aggregation relationship of the geographic area, a network-layer routing strategy is designed to enhance routing scalability. Compatibility is improved by deploying the network-layer designs only at the network edge where granularity-customizable geocast is implemented, without requiring changes to the current IP infrastructure. Then, the network-layer functions are integrated with an application-layer mapping service to support intercommunication between different network edges. Furthermore, a prototype system is built to implement and evaluate the proposed L3Geocast, which outperforms the existing approaches in terms of communication latency and mapping overhead.

Task cooperation emerges as a efficacious strategy for the execution of intricate tasks within the context of challenging wireless edge networks characterized by limited resources and intermittent infrastructure connections. Presently, TCP/IP-based solutions encounter issues related to suboptimal utilization of network resources and a substantial dependency on infrastructure connections. In light of this, Information-Centric Networking (ICN) has surfaced as a promising architectural paradigm aimed at mitigating these challenges. In ICN-based task cooperation, the data reuse characteristic of ICN enhances the efficiency of network resource utilization. However, this also introduces plausible security vulnerabilities to the reused data, encompassing eavesdropping attacks and unauthorized access attacks. In this paper, we propose an ICN-based secure task cooperation scheme to mitigate the above threats without compromising the efficiency of task execution. We present the task cooperation model that quantifies the cost of securing data reuse in task execution. We also introduce a specific naming convention to support the acquisition of collaborative task requests and keys related to the task cooperation. Besides, we introduce a novel design for an enhanced name-based access control scheme that ensure both data confidentiality and access control in collaborative tasks accessed by the same sub-policy, streamlining the encryption process for content keys. Security analysis and experimental results demonstrate that our scheme effectively safeguards the security of data reuse. Furthermore, compared to existing schemes, our scheme incurs lower cost associated with computation and security.

Task cooperation is an effective way to execute a complex task in challenging wireless edge networks. Existing TCP/IP-based solutions encounter the problem of low network resource utilization and the heavy dependency of infrastructure connections. Information-centric networking(ICN) is a promising architecture to address these issues. In existing ICN-based task cooperation schemes, the data reuse feature of ICN improves the utilization of network resources, which also brings potential security threats to the reused data. To guarantee the security of data reuse in task cooperation without affecting the data reuse feature, we propose an ICN-based secure task cooperation scheme. In our scheme, the specific naming convention is designed to support task cooperation and the acquisition of keys. In addition, our scheme implements fine-grained access control for data reuse in task cooperation combined with attribute-based encryption. Experimental results show that our scheme enhances the security of task cooperation with low cost compared with existing schemes.