This paper extends a framework for load balancing in large Wi-Fi networks we previously proposed. The framework partitions the network into clusters and assigns a sub-controller to each cluster. The sub-controllers employ a Deep Q-Learning-based algorithm to balance the load on the access points in the cluster. The sub-controllers collaborate by exchanging their training updates through a database. However, in highly dynamic Wi-Fi networks, the frequency of these exchanges may cause control plane overhead. In this paper, the training updates are uploaded only when their “quality” meets a condition dependent on a threshold. High threshold values reduce overhead but compromise local learning performance, and vice versa. Therefore, finding the optimal threshold values is formulated as a multi-objective optimization problem. Two metrics are designed to quantify the overhead and instability. These metrics are stochastic, scenario-dependent, and time-consuming to estimate. Randomly generated scenarios of large and dynamic 802.11ax networks are simulated to collect realizations of the metrics. The objectives are obtained by smoothing the realizations with Gaussian kernel regression. The problem is then solved with a genetic algorithm to estimate the Pareto-optimal threshold values.

Object detection has recently seen an interesting trend in terms of the most innovative research work, this task being of particular importance in the field of remote sensing, given the consistency of these images in terms of geographical coverage and the objects present. Furthermore, Deep Learning (DL) models, in particular those based on Transformers, are especially relevant for visual computing tasks in general, and target detection in particular. Thus, the present work proposes an application of Deformable-DETR model, a specific architecture using deformable attention mechanisms, on remote sensing images in two different modes, especially optical and Synthetic Aperture Radar (SAR). To achieve this objective, two datasets are used, one optical, which is Pleiades Aircraft dataset, and the other SAR, in particular SAR Ship Detection Dataset (SSDD). The results of a 10-fold stratified validation showed that the proposed model performed particularly well, obtaining an F1 score of 95.12% for the optical dataset and 94.54% for SSDD, while comparing these results with several models detections, especially those based on CNNs and transformers, as well as those specifically designed to detect different object classes in remote sensing images.

The growing complexity of cyber incidents presents significant challenges for digital forensic investigators, especially in evidence collection and analysis. Public resources are still limited because of ethical, legal, and privacy concerns, even though realistic datasets are necessary to support research and tool developments. To address this gap, we introduce ForensicsData, an extensive Question-Context-Answer (Q-CA) dataset sourced from actual malware analysis reports. It consists of more than 5,000 Q-C-A triplets. A unique workflow was used to create the dataset, which extracts structured data, uses large language models (LLMs) to transform it into Q-C-A format, and then uses a specialized evaluation process to confirm its quality. Among the models evaluated, Gemini 2 Flash demonstrated the best performance in aligning generated content with forensic terminology. ForensicsData aims to advance digital forensics by enabling reproducible experiments and fostering collaboration within the research community.

The increasing complexity of modern WiFi networks aligns them more closely with cellular systems. This convergence underscores the need for WiFi-specific LLMs, akin to ongoing efforts in 5 G. An essential initial step is the design of a WiFi dataset compatible with LLM requirements, structurally coherent, and containing both technical and general information to ensure broad applicability. This work introduces WiFiQnA, a curated dataset of WiFi-related multiple-choice questions designed for LLM fine-tuning. We define two multiple-choice question (MCQ) formats: general knowledge and procedural configuration/troubleshooting questions. We develop a multi-step generation framework using three LLMs for question generation and four for validation. The process integrates filtered telecom datasets, WiFi-specific sources, and tailored prompts, ensuring semantic diversity, accuracy, and reliability.

Federated Learning (FL) is a machine learning technique that uses a decentralized approach to train a model across several servers or devices. In contrast to centralized machine learning approaches, where all nodes or devices share data at a central location for training, FL implies that each device retains its data locally and does not share it with the server. In FL, the server functions as a central coordinator that gathers model parameters and updates from individual client devices, combines them into a unified model, and redistributes the updated model to the clients. Our research involved comparing different federated lerning approaches for detecting malware in network traffic. To implement and train the machine learning model, we utilized the Flower framework. This research focused on comparing several aggregation methods: FedAvg, FedProx, and FedYogi. These aggregators were applied to the UNSW-NB15 dataset. Among them, FedProx produced the best results.

Developing efficient load-balancing techniques remains a persistent research challenge as modern WiFi networks evolve into increasingly complex environments, incorporating new enhancements in their standards. For instance, DQL-MultiMDP is a load-balancing algorithm that learns an optimal STA-to-AP association policy to ensure user fairness and optimize network performance in dense and dynamic WiFi networks. The algorithm leverages a Multi-Markov Decision Process (MultiMDP) strategy to accommodate the fluctuating number of devices caused by their switching on/off. However, scalability challenges arise due to the exponential expansion of the action space. In this paper, we propose a divide-and-conquer approach that extends the algorithm to operate in extremely large deployments: a dynamic partitioning mechanism divides the network into clusters and assigns a sub-controller to manage the STA-to-AP association in each cluster independently, and a coordination mechanism enables them to exchange their training updates. Experimental investigations validate the effectiveness of the approach and motivate future work.

In this paper, we present our primary version of DQL-MultiMDP, a flexible load balancing approach designed to ensure long-term user satisfaction in modern WiFi networks (WiFi 7 and beyond). The algorithm quasi-simultaneously solves multiple Markov Decision Processes (MDPs) by switching between them depending on the environmental configu-ration, specifically the number of Access Points (APs) and user Stations (STAs). Leveraging Deep Q-Learning (DQL), it utilizes sophisticated state representations with advanced metrics to determine the optimal AP-to-STA association through autonomous decision-making. Experimental results validate the effectiveness of the approach and lay the foundation for further improvements.

Energy Efficiency (EE) is a key performance metric to design future wireless networks. Since Directional Antennas (DAs) focus the transmission energy towards the destination, it has been shown as a cost-effective solution when used in a backhaul network. In this paper we propose a new joint optimization framework of energy consumption and throughput in backhaul Wireless Mesh Networks (WMNs) equipped with DAs. We first formulate the joint optimization problem as a Mixed Integer Linear Problem (MILP) using a weighted objective function of both the consumed energy and the throughput. Then, we propose to use the Ant-Q algorithm, a Reinforcement Learning (RL) based approach, to reduce the solution complexity and enhance its convergence. Considering a discrete power control scheme, we define a new routing scheme based on the Ant-Q heuristic to select jointly the transmission beam and the transmission power. Using ILOG Cplex to find the optimal solution and NS-3 to conduct extensive simulations, we show the effectiveness and the accuracy of the proposed routing algorithm. Moreover, we analyze the optimization tradeoff depending on the beamwidth, the network topology, the gateway position and the optimization weight factor.

We propose a model based on stochastic geometry to assess downlink interference and signal over interference ratio (SIR) in LTE networks. The originality of this work lies in the proposition and combination of resource blocks assignment strategies, transmission power control, and realistic traffic patterns into a stochastic geometry model. For this model, we compute the first two moments of interference. They are used to parameterize its distribution from which we deduce the SIR distribution. Outage and transmission rates (modulation and coding rate) are then derived to evaluate the system performance. Simulations that cover a large set of scenarios show the accuracy of our proposal and allow us to compare these strategies with more complex ones that aim to minimize global interference. Numerical evaluations highlight the behavior of the LTE network for different traffic patterns/load, eNodeB density, and amount of resource blocks and offer insights about possible parameterization of LTE networks.

Directional Antennas (DAs) provide higher gain, and reduce interference by directing beams toward the desired receiver. In this paper, we propose a new joint optimization framework considering the energy consumption and throughput in DAs Wireless Mesh Networks (WMNs). We formulate the joint optimization problem as a Mixed Integer Linear Problem (MILP) using a weighted objective function of both the consumed energy and the throughput. We use ILOG Cplex [1], a software based on branch and cut method, to find the optimal solution of the optimization problem. Results prove the efficiency of using DAs in WMNs in the considered scenarios and show that the consumed energy increases with the beamwidth and decreases with the number of power levels when using power control.

Energy Efficiency is an important feature in poor-covered areas where not only the access to a cellular network is scarce but also energy sources are limited. In this paper, we consider a wireless mesh network to act as a local backhaul network to cover rural and remote villages. The shortage of energy in these settlements motivates to optimize the energy consumption of the wireless backhaul network. Therefore, we propose to use Directional Antennas (DAs) to improve the throughput and the network energy consumption. DAs focus the RF signals toward the desired destination, to reduce the collisions and the number of hops between the source and the destination. We provide for both OAs and DAs networks, the number of links (hops) that a packet passes through to reach the destination. Using extensive simulation, we evaluate the network performance in terms of packet loss, mean throughput, mean energy consumption and energy efficiency for the chain and grid topologies. Simulations results show that using DAs improves the throughput and the energy efficiency, and reduce the mean loss ratio, and the consumed energy.

A 1-tier network composed by multimode low power nodes (LTE/Wifi) is considered as a cost-efficient solution for operators to improve services in poorly or uncovered rural areas. Using an interference coordination technique, network performance can be further improved. Stochastic geometry gives a set of tools to model the location of base stations and user equipments in such wireless networks. Using a spatial model we analyse the network performance in terms of coverage probability and data rate. To realistically model multimode node locations, a new point process model, called the r-l square point process (p.p.), is used in this work. The model of downlink communication including the coordination technique is developed and it allows to evaluate the system performances in term of coverage probability and throughput. Results show that cooperation among nodes improves the network performance.

In this paper, we propose to use the Joint Transmission approach of Coordinated Multipoint (JT-COMP) of cellular networks to reduce the interference in dense Very High Throughput (VHT) wireless LANs. VHT WLANs are based on wider channel bandwidth, efficient modulation techniques and support for spatial streams using MIMO schemes. However, the interference problem persists despite these approaches, and thereby prevents mobile stations from fully reaping the capacity improvement of such networks. In order to optimize the coverage and minimize the cell overlap in dense stadium scenario, AP locations must be planned carefully. To this end, we model positions of nodes using a spatial stochastic model called the r-l square point process. Then, we derive the coverage probability and throughput expressions and investigate the benefit of Joint Transmission coordination technique. Using simulation, we characterize the performance metrics for different sizes of coordinated set and carrier sensing domain of access points. Our results show that JT-CoMP is a promising scheme for dense WLANs.