In vehicular networks, decentralized cooperative learning strategies have gained significant attention due to the lower communication overhead they involve when compared to centralized cooperative learning approaches like Federated Learning. Decentralized solutions enable vehicles to collaboratively train Machine Learning (ML) models by exchanging parameters without relying on a central server. However, conventional model-sharing methods still suffer from high communication overhead and increased vulnerability to poisoning attacks. This paper presents SPARE , a gossip-based cooperative learning protocol that leverages Vehicle-to-Vehicle (V2V) communication to enhance communication efficiency by exchanging selected model parameters. SPARE selects vehicle nodes for model updates and transmits only the most significantly updated layers, reducing redundancy and improving efficiency. This selective exchange minimizes communication resource consumption and enhances privacy, as the complete model is never shared across the network. We assess the proposed approach using a real-world driving dataset, featuring data from multiple drivers along the same route. Experimental results prove that our method achieves efficient learning with significantly lower communication overhead, demonstrating its suitability for deployment in resource-constrained vehicular networks.

According to the Ethicnet paradigm, which has been recently proposed, communication networks are called to consider Key Value Indicators (KVI)s besides the traditional Key Performance Indicators (KPI)s. The actual realization of the Ethicnet, however, requires several issues to be addressed. One of the most important is the identification of the KVIs to be considered and the way in which they can be measured, controlled, and monitored. This paper contributes to the discussion regarding such topic by answering fundamental questions regarding the identification of the KVIs, the measurement of the extent to which a given network solution contributes to the KVIs, and the definition of requirements in terms KVIs.

In this work i-WSN League, a comprehensive hardware/software framework for the support of distributed training and inference is introduced. For what concerns the hardware, in i-WSN League two types of nodes are considered, namely, head nodes and common nodes. Head nodes are resource rich nodes that have the capabilities for training artificial neural network. Common nodes collect data and can execute inference only. In i-WSN League, all nodes are grouped in Clusters, each with a Cluster Head (selected among the head nodes), which is the only node responsible for training. To this end, data coming from all nodes in the Cluster can be utilized. This, however, involves large exchange of data which might be unsustainable by common nodes. Thus, only part of the data collected by common nodes is sent to the Cluster Heads and a network of Cluster Heads will implement distributed learning in a peer-to-peer fashion. As compared to state of the art literature, the key contributions of our work are related to the combination of gossiping and clustering to adapt the operations executed by each node to its capabilities, with the aim of minimizing the energy consumption in resource limited nodes, while preserving accuracy. In this paper i-WSN League is assessed in a simple scenario in which a wireless sensor network monitors the air pollution in a large city. Performance results obtained by considering auto-encoders prove the effectiveness of the proposed scheme as well as its balanced energy consumption and fairness in resource consumption distribution.

In this paper, a software defined networking approach, named SDN-(UAV)ISE, is introduced for wireless sensor networks with data mules. In our scenario, sensor nodes are equipped with two wireless interfaces: one exploits a long range, low data rate wireless technology, such as LoRa or Sigfox, whereas the other uses a short range, wireless technology that provides higher data rate, examples include IEEE 802.15 or IEEE 802.11. Sensors can communicate directly to a control center through the long range wireless interface, or by utilizing the multihop paradigm realized with the aid of the short range wireless communication interface. In this context, a drone acts as a mobile sink (i.e., the mule) for the latter case. The movement of the data mule is forecast by the SDN controller and the forecast positions are considered to generate the flow table entries to be installed in the sensor nodes and schedule their applications. To this purpose, it is expected that the drone will move to the locations where abnormal conditions are observed by the sensors. In our work, a simple and efficient decision tree algorithm is implemented, which takes the values measured by the sensors as inputs, to forecast the route of the data mule. The proposed scheme is assessed through a large experimental campaign considering different operating conditions.