Abdellah EL Mekki, Samar M. Magdy, Houdaifa Atou, Ruwa AbuHweidi, Baraah Qawasmeh, Omer Nacar, Thikra Al-hibiri, Razan Saadie, Hamzah A. Alsayadi, Nadia Ghezaiel Hammouda, Alshima Mohammed Alkhazimi, Aya Hamod, Al-Yas Yaqoob Al-Ghafri, Wesam El-Sayed, Asila Ismail al Sharji, Mohamad Ballout, Anas Belfathi, Karim Ghaddar, Serry Sibaee, Alaa Aoun, Aeej Mohammed Aseri, Lina Abureesh, Ahlam Bashiti, Majdal Yousef, Abdulaziz Hafiz, Yehdih Mohamed, Emira Hamedtou, Brakehe Emehah, Rahaf Alhamouri, Youssef Nafea, Aya El Aatar, Walid Al-Dhabyani, Emhemed S. Hamed, Sara Shatnawi, Fakhraddin Alwajih, Khalid Elkhidir, Ashwag Alasmari, Abdurrahman Gerrio, Omar Said Alshahri, AbdelRahim A. Elmadany, Ismail Berrada, Amir Azad Adli Al-kathiri, Fadi Zaraket, Mustafa Jarrar, Yahya Mohamed EL Hadj, Hassan Alhuzali, Muhammad Abdul-Mageed. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Recent advances in text-to-SQL models, which translate natural language questions (NLQs) into executable SQL queries, have made interacting with relational databases more accessible, even for those with limited technical ability. This task has seen significant improvement with the release of multiple English datasets and benchmarks such as WikiSQL, SPIDER, and BIRD, each covering different domains and levels of complexity. Non-English high-resource languages, such as Chinese, Russian, and Arabic, have also benefited from these advances, either through the translation of existing datasets or the creation of new ones. Dialect2SQL , a newly released text-to-SQL dataset, is dedicated to the Moroccan dialect (Darija), which is known for its complexity and distinctiveness compared to other Arabic dialects and Modern Standard Arabic. In this article, we conduct a comprehensive study on text-to-SQL for Darija by conducting several experiments mainly on the Dialect2SQL dataset using different approaches and configurations with two code-based large language models, StarCoder2 and Qwen-2.5-Coder. The experiments reveal the performance gap between models fine-tuned on English data and those fine-tuned on Darija. Additionally, the results illustrate the positive impact of incorporating multi-language datasets during training. In particular, the gap decreases from 10.1% to 6.7% in BLEU, and from 12.5% to 5.7% in TSED.

Emotion recognition plays an important role in enhancing social communications and is a key component of improving human-machine interactions. With the rise of deep learning, emotion recognition systems are becoming more effective and accurate. However, progress is often limited by the availability of datasets, a problem affecting low-resourced languages such as the Moroccan Arabic Dialect (Darija). This article introduces the Moroccan Arabic Multimodal Authentic Emotional Dataset (MORAD), a multimodal dataset collected to address the scarcity of affective datasets in the Moroccan Arabic Dialect. MORAD comprises 2325 videos, totaling 3 h, and includes 329 unique speakers expressing their emotions naturally in unscripted real-world settings. The dataset contains seven emotions: 1) happiness; 2) sadness; 3) anger; 4) fear; 5) disgust; 6) surprise; and 7) neutrality. We evaluated various deep learning models on this dataset, including unimodal speech, visual, and textual models, as well as a multimodal approach. The multimodal model achieved the highest performance with a weighted F1 score of 73.11% and an accuracy of 73.36% on a 4-way classification task, and 66.69% and 69.38%, respectively, on the 7-way task. These results provide a benchmark demonstrating MORAD’s utility for emotion recognition.

As software systems increasingly rely on natural language interfaces, ensuring the reliability of these systems is crucial. One critical component is the ability to accurately translate natural language queries into corresponding SQL queries, a field known as Text-to-SQL. However, the scarcity of high-quality, large-scale, and domain-specific Text-to-SQL datasets hinders the development of reliable and robust models. To tackle these challenges, we propose SelectCraft , a novel automatic generation approach designed to create realistic Text-to-SQL datasets tailored to specific domains. Our method leverages existing databases and their structures to generate complex text-SQL pairs that mirror real-world usage scenarios. As a proof of concept, we have successfully generated a substantial financial Text-to-SQL dataset, denominated as BanQies , encompassing over 1 million samples utilizing our proposed approach. Moreover, we introduce BanQL , a new large language model (LLM) based on StarCoder2 , a state-of-the-art code-based LLM, and fine-tuned on our newly created dataset. We evaluate BanQL performance against several state-of-the-art models, demonstrating significant enhancements in accuracy and generalizability, highlighting the advantages of incorporating domain-specific data in Text-to-SQL tasks. We firmly believe that our contributions have the potential to improve the overall reliability of Text-to-SQL software systems. CCS Concepts: • Software and its engineering; • Computing methodologies → Natural languageprocessing; • Information systems → Structured Query Language;

Fakhraddin Alwajih, Abdellah El Mekki, Samar Mohamed Magdy, AbdelRahim A. Elmadany, Omer Nacar, El Moatez Billah Nagoudi, Reem Abdel-Salam, Hanin Atwany, Youssef Nafea, Abdulfattah Mohammed Yahya, Rahaf Alhamouri, Hamzah A. Alsayadi, Hiba Zayed, Sara Shatnawi, Serry Sibaee, Yasir Ech-chammakhy, Walid Al-Dhabyani, Marwa Mohamed Ali, Imen Jarraya, Ahmed Oumar El-Shangiti, Aisha Alraeesi, Mohammed Anwar AL-Ghrawi, Abdulrahman S. Al-Batati, Elgizouli Mohamed, Noha Taha Elgindi, Muhammed Saeed, Houdaifa Atou, Issam Ait Yahia, Abdelhak Bouayad, Mohammed Machrouh, Amal Makouar, Dania Alkawi, Mukhtar Mohamed, Safaa Taher Abdelfadil, Amine Ziad Ounnoughene, Anfel Rouabhia, Rwaa Assi, Ahmed Sorkatti, Mohamedou Cheikh Tourad, Anis Koubaa, Ismail Berrada, Mustafa Jarrar, Shady Shehata, Muhammad Abdul-Mageed. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Deep learning models for time series forecasting increasingly face a critical trade-off between accuracy and computational efficiency especially in mobile, wireless, and edge computing scenarios. While state-of-the-art architectures like TimeMixer achieve strong performance through multi-scale decomposition, their large parameter footprints limit real-world deployment. In this paper, we propose Efficient Hybrid TimeMixer, a novel architecture that combines TimeMixer's decompositional strengths with lightweight, conditionally-applied enhancements derived from TimesNet. Our approach introduces three key innovations: (1) conditional processing to apply computationally intensive modules only when necessary; (2) parameter-efficient fusion mechanisms to integrate components without overhead; and (3) adaptive scale processing to allocate resources based on temporal complexity. Extensive experiments on five benchmark datasets show our model achieves up to$\mathbf{1 4. 5 6 \%}$fewer parameters while maintaining or surpassing eleven competitive baselines forecasting accuracy. Specifically, on the ETTm2 dataset with a 96-step horizon, our model surpasses TimeMixer's using only$\mathbf{2. 0 0 M}$parameters.

Bashar Talafha, Karima Kadaoui, Samar Mohamed Magdy, Mariem Habiboullah, Chafei Mohamed Chafei, Ahmed Oumar El-Shangiti, Hiba Zayed, Mohamedou Cheikh Tourad, Rahaf Alhamouri, Rwaa Assi, Aisha Alraeesi, Hour Mohamed, Fakhraddin Alwajih, Abdelrahman Mohamed, Abdellah El Mekki, El Moatez Billah Nagoudi, Benelhadj Djelloul Mama Saadia, Hamzah A. Alsayadi, Walid Al-Dhabyani, Sara Shatnawi, Yasir Ech-chammakhy, Amal Makouar, Yousra Berrachedi, Mustafa Jarrar, Shady Shehata, Ismail Berrada, Muhammad Abdul-Mageed. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

This paper examines the actions taken by commercial banks to develop effective surveillance strategies for the identification of the origins of clients' digital assets within the cryptocurrency sector. Informed by the European Union Agency for Cybersecurity (ENISA) guidelines, this study integrates insights from compliance expert interviews, a case study of Barclays, and a thorough analysis of existing literature to enhance cybersecurity measures in response to increasing threats in digital transactions such as cryptojacking. The research emphasizes constructing a methodical approach for organizations to enhance their established defenses against cryptocurrency-related risks by integrating ENISA principles. The proposed strategy focuses on improving transaction traceability and monitoring capabilities of blockchain technologies while acknowledging technological challenges related to regulatory and safety compliance. The provided in-depth analysis of blockchain and cryptocurrency technologies present in this paper advocates for strict regulatory frameworks to navigate these complexities and enhance overall resilience towards this threat.

Abdellah El Mekki, Muhammad Abdul-Mageed, ElMoatez Billah Nagoudi, Ismail Berrada, Ahmed Khoumsi. Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Prediction and analysis of the severity of road traffic accidents have become a top priority and concern for road traffic safety. Reducing the accident risk and traffic congestion caused by accidents is still one of the most critical issues in the intelligent transportation system (ITS). This paper presents an effort to analyze the factors that influence the severity of accidents with different environmental experiences. This study is based on a public dataset of 2.25 million cases of car traffic accidents in the United States. Methodologically, we have adopted gradient boosting algorithms (i.g. XGBoost, LightGBM, and CatBoost), which have been recognized as robust classifiers, especially in handling feature values of categorical variables. In addition, we employed the SHapley Additive explanation (SHAP) technique to understand the black box outputs. This study identifies risk factors, including weather, time, spatial, and road analysis. Also, a detailed interpretation of models output was considered to understand geospatial safety effects at the micro, meso, and macro levels. The experiment results suggest that the importance of risk factors varies across levels in different severities. These results filled research gaps by conducting a longitudinal analysis of environmental factors in accident severity. Meanwhile, it can provide helpful insights for personal travelers and government agencies.

With the ever-increasing network intrusion techniques, the effectiveness of conventional Network Intrusion Detection Systems (NIDS) solutions has become limited. As a response, machine learning-based NIDS have emerged as a potential alternative to handle newly designed intrusions. However, developing NIDS using machine learning methods necessitates high-quality labeled datasets, which can be time-consuming and resource-intensive. In this paper, we propose and assess a domain adaptation method specifically designed for NIDS. Initially, we represent flows as images using raw packets and NFStream. Next, we utilize convolutional neural networks to extract relevant features and employ gradient reversal for domain adaptation. This allows us to leverage a labeled dataset (source domain) to construct models that perform well on an unlabeled dataset (target domain). To assess the performance of our approach, we leverage suitable evaluation metrics and three publicly available datasets, namely USTC-TC2016, CIC-IDS2017, and CUPID. The results obtained indicate the need for further investigation into domain adaptation techniques for NIDS, potentially leading to improved intrusion detection capabilities.

In this work, CT-xCOV, an explainable framework for COVID-19 diagnosis using Deep Learning (DL) on CT-scans is developed. CT-xCOV adopts an end-to-end approach from lung segmentation to COVID-19 detection and explanations of the detection model’s prediction. For lung segmentation, we used the well-known U-Net model. For COVID-19 detection, we compared three different CNN architectures: a standard CNN, ResNet50, and DenseNet121. After the detection, visual and textual explanations are provided. For visual explanations, we applied three different XAI techniques, namely, Grad-Cam, Integrated Gradient (IG), and LIME. Textual explanations are added by computing the percentage of infection by lungs. To assess the performance of the used XAI techniques, we propose a ground-truth-based evaluation method, measuring the similarity between the visualization outputs and the ground-truth infections. The performed experiments show that the applied DL models achieved good results. The U-Net segmentation model achieved a high Dice coefficient (98%). The performance of our proposed classification model (standard CNN) was validated using 5-fold cross-validation (acc of 98.40% and f1-score 98.23%). Lastly, the results of the comparison of XAI techniques show that Grad-Cam gives the best explanations compared to LIME and IG, by achieving a Dice coefficient of 55%, on COVID-19 positive scans, compared to 29% and 24% obtained by IG and LIME respectively. The code and the dataset used in this paper are available in the GitHub repository [1].

The rising frequency and complexity of cyber threats have necessitated the development of effective Network Intrusion Detection Systems (NIDS). Anomaly-based detection approaches have gained prominence for their ability to detect unknown and sophisticated attacks. In this paper, we introduce NF-NIDS, a novel approach for anomaly-based network intrusion detection using Normalizing Flows (NFs) to accurately classify network traffic into normal or malicious categories given the assumption of the availability of scarce attack samples. To address the challenge of limited attack samples, we employ two flow-based models, namely Inverse Autoregressive Flow (IAF) and Neural Spline Flows (NSF). These models are used to learn the underlying distribution of normal traffic and generate pseudo-attacks from the tails of the distribution. To evaluate the effectiveness of NF-NIDS, we conducted experiments on three well-known network datasets. The results demonstrate that our approach achieves high performance levels while incurring low to negligible false discovery rates. Specifically, NF-NIDS yields impressive results, with an accuracy of 98.71% for USTC-TFC2016, 94.86% for CIC-IDS2017, and 98.20% for CIC-IDS2018. In terms of the F1-score, NF-NIDS scores 98.72% for USTC-TFC2016, 97.05% for CIC-IDS2017, and 99.51% for CIC-IDS2018.

The extraction of text information from traffic panels is one of the challenging problems in computer vision. Although the past decade has seen a promising shift and important progress in object detection, few works and a limited number of datasets focus specifically on extracting text from traffic signs. To address the lack of data for text detection in traffic panels, especially those with Arabic scripts, this paper introduces a new multilingual and multipurpose dataset named ASAYAR. It consists of three sub-datasets: Arabic-Latin scene text localization, traffic sign detection, and directional symbol detection. The dataset contains 1763 images collected on different Moroccan highways, and annotated manually, using 16 object categories. The fully annotated ASAYAR images contains more than 20000 bounding box objects. The paper also investigates the usability of the dataset, by evaluating the performance of thestate-of-the-artalgorithms for object and text detection. Experimental results show good detection scores, demonstrating the potential contribution of ASAYAR in the development of methods for text extraction from traffic panels.

Worldwide, traffic accidents are recognized as one of the leading causes of death. This phenomenon leads to significant daily losses affecting both road users and road authorities. Therefore, the need for effective dynamic road security systems is highly considered. Traffic accident data analysis is one of the promising approaches for improving road safety. By taking into account multiple factors (e.g., infrastructure, weather, driver behavior, etc.), it allows measuring the impact of traffic accidents on road security. However, reformulating this impact into practical road safety decisions remains limited and unstructured. To overcome the mentioned limitations, this paper proposes the first end-to-end recommendation framework for road safety. Our framework introduces a three-layered architecture, designed to handle data analysis and action recommendation tasks. For data analysis, we adopt a baseline of state-of-the-art machine and deep learning algorithms to build different traffic accident prediction models. For the action recommendation task, we developed a new approach involving model predictions, model interpretations, actions definition, and road-action interactions matrix annotation. The proposed framework has been successfully experimented and evaluated using two real-world datasets of historical traffic accidents of France (2006-2017) and Morocco (2010-2014), achieving interesting ROC-AUC scores of 0.93 and 0.96, respectively.

The emergence of Internet of Vehicles (IoV), as a large-scale distributed system, introduces new research and development challenges for supporting resource-constrained devices. In fact, the latency of retrieving contents and performing the desired tasks may increase dramatically and failures may occur when resource limits are exceeded. In this paper, we assess the use of advanced machine learning paradigms to achieve more accurate personalized edge caching and replacement decisions, while supporting data privacy, vehicle mobility, time-varying and location-aware content popularity. Firstly, we show that popular federated learning-based schemes fail in maintaining acceptable performance under the above settings. Secondly, we propose a scalable-by-design edge caching scheme for IoV, leveraging decentralized region-to-region Road Side Units (RSUs) exchanges while enhancing region local models. Finally, simulation results show that our scheme achieves higher performance in terms of average delay and edge hit ratio while keeping the cost and privacy risks at a minimum.

Federated learning (FL) is a well-established distributed machine-learning paradigm that enables training global models on massively distributed data i.e., training on multi-owner data. However, classic FL algorithms, such as Federated Averaging (FedAvg), generally underperform when faced with Non-Independent and Identically Distributed (Non-IID) data. Such a problem is aggravated for some hyperparametric methods such as optimizers, regularization, and normalization techniques. In this paper, we introduce FedBS, a new efficient strategy to handle global models having batch normalization layers, in the presence of Non-IID data. FedBS modifies FedAvg by introducing a new aggregation rule at the server-side, while also retaining full compatibility with Batch Normalization (BN). Through our evaluations, we have empirically proven that FedBS outperforms both classical FedAvg, as well as the state-of-the-art FedProx through a comprehensive set of experiments conducted on Cifar-10, Mnist, and Fashion-Mnist datasets under various Non-IID data settings. Furthermore, we observed that in some cases, FedBS can be 2× faster than other FL approaches, coupled with higher testing accuracy.

Abdellah El Mekki, Abdelkader El Mahdaouy, Ismail Berrada, Ahmed Khoumsi. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

Statistics show that the global number of cars on the road will nearly double by the year 2040. The widespread use of cars prompts the search for new technologies and systems to ensure road safety, such as driving assistance systems, driver monitoring devices and driver training programs. For a number of these systems, driving maneuver recognition is a core function indispensable for correct operation. This paper addresses the problem of driving maneuver detection using smartphone sensors, especially accelerometers and gyroscopes. A framework based on a number of deep learning methods for maneuver classification and clustering was introduced. We studied 13 types of maneuvers. Three classifiers, each achieving good performance for recognizing the considered set of events, were selected, and their combination into an optimal set of classifiers was investigated. Our approach was tested on a real-world dataset, and achieved a good detection rate for 7 maneuvers with a balanced accuracy of 0.90 and an average F1 score of 0.71, which outperforms the other state-of-the-art recognition systems.

Cut-in into vehicle platoons is a dangerous driving maneuver that affects the safety and efficiency of platooning vehicles. An accurate prediction of such maneuver enables the platooning system to take safety measures that ensure the platoon safety and integrity. The contribution of this paper consists of an evaluation of a set of supervised machine learning algorithms for cut-in maneuver recognition, for eventual use in platooning systems. The models were trained and tested on a large-scale publicly available driving dataset, from which cut-in events were extracted. The results show that tree-based classifiers such as Gradient Booting Machine can recognize the cut-in maneuvers with an Fl-score of 98%. An experiment to investigate the model performance with advanced prediction times shows that up to 80.5% of the cut-ins were correctly predicted 1 second before the lane crossing time.

Traffic light detection and classification represent a major issue for autonomous driving. Although a number of works have been published on this topic, providing a real-time processing solution is still a challenging task. In this paper, we show, by experimenting three models, namely “Faster R-CNN”, “R-FCN” and “SSD” on and two datasets, namely “Bosch Small Traffic Light Dataset” and “Lisa Traffic Light Dataset”, that we can achieve a higher accuracy while reducing the detection and recognition time. In order to improve the overall performance and take the best score of the trained models, we used the ensembling modeling technique. The obtained results outperform the state-of-the-art.

In this paper, we will present our project concerning realizing a system of predicting maneuvers before the vehicle turns through using a dataset containing videos of drivers in different situations and different maneuvers, as well as information on the environment, such as speed, empty lines and the existence of an artifact. After preparing the dataset, we built our CNN-LSTM model based on convolutional and recurrent layers. Our CNN-LSTM model allowed us to predict the maneuver with an accuracy of 94.1% and 3.75 seconds before the turn. Finally, For our model to be robust, we tried to detect anomalies and replace them with more meaningful values. We also tested our model by adding noise to the images.

The recent computerizations of cars, together with the development of sensor technologies and car communication devices have transformed the cars into wealthy sources of information. The analysis of data generated continuously by cars can contribute greatly in improving driving safety and drivers comfort. Even though different analytical solutions have emerged recently, there still exist some important issues in driving safety that we assume were poorly addressed, as well as diverse mathematical methodologies whose application in driving behavior analysis is to be investigated. In this paper, we developed a methodology to process and analyze car-generated data, with focus on two analysis goals: 1) automatic verification of drivers' behavior conformity to traffic rules; and 2) visualization and comparison of drivers' behaviors. The proposed methodology is divided into three steps. At first, the abstraction using numerical domains is used to reduce the size of the generated data. Then, the probabilistic graphical models (Probabilistic Automata, and Labeled Directed Graphs) combined with a machine-learning algorithm are used for building a formal model of the driver behavior. Finally, two indepth analyses are carried out by applying automatic model checking and graph matching techniques. Early experimental results point out that the design of numerical domains considered influences hugely the analysis results.

Traffic sign recognition (TSR) represents an important feature of advanced driver assistance systems, contributing to the safety of the drivers, pedestrians and vehicles as well. Developing TSR systems requires the use of computer vision techniques, which could be considered fundamental in the field of pattern recognition in general. Despite all the previous works and research that has been achieved, traffic sign detection and recognition still remain a very challenging problem, precisely if we want to provide a real time processing solution. In this paper, we present a comparative and analytical study of the two major approaches for traffic sign detection and recognition. The first approach is based on the color segmentation technique and convolutional neural networks (C-CNN), while the second one is based on the fast region-based convolutional neural networks approach (Fast R-CNN).

This paper deals with modeling human behavior routines during driving. We propose a new vision of the maximum causal entropy framework for inverse reinforcement learning to predict actions to be triggered in particular situation (lane change). We designed a plugin to enhance functionalities of the vCar platform which is presents an open source solution for the analysis and visualization of data from cars [7].

The international conference on wireless networks and mobile communications (WINCOM'16) is held in the imperial city and Scientific capital of Morocco, Fez, on 26–29 October 2016, Hotel Medina Palace.

Abstract The core of cognitive radio paradigm is to introduce cognitive devices able to opportunistically access the licensed radio bands. The coexistence of licensed and unlicensed users prescribes an effective spectrum hole‐detection and a non‐interfering sharing of those frequencies. Collaborative resource allocation and spectrum information exchange are required but often costly in terms of energy and delay. In this paper, each secondary user (SU) can achieve spectrum sensing and data transmission through a coalitional game‐based mechanism. SUs are called upon to report their sensing results to the elected coalition head, which properly decides on the channel state and the transmitter in each time slot according to a proposed algorithm. The goal of this paper is to provide a more holistic view on the spectrum and enhance the cognitive system performance through SUs behavior analysis. We formulate the problem as a coalitional game in partition form with non‐transferable utility, and we investigate on the impact of both coalition formation and the combining reports costs. We discuss the Nash Equilibrium solution for our coalitional game and propose a distributed strategic learning algorithm to illustrate a concrete case of coalition formation and the SUs competitive and cooperative behaviors inter‐coalitions and intra‐coalitions. We show through simulations that cognitive network performances, the energy consumption and transmission delay, improve evidently with the proposed scheme. Copyright © 2016 John Wiley & Sons, Ltd.

Driver behavior has long been considered as particularly relevant for the development of automotive applications, especially that recently these applications are increasingly trying to adapt to the driver. However, drivers behave differently in the different traffic situations, hence the need of techniques to enable cars to learn from their drivers and create a model of his behavior. Actually, future generation of cars will be equipped with all sorts of sensing, computing and communication devices that will allow them to acquire all information about the state of the vehicle, the driver and the environment. And, hence make easier the driving behavior learning process. The present paper addresses the problem of modeling and learning the behavior of a driver in an intelligent car by presenting an approach for the construction and verification of a learned driving model. First, we propose a new way for modeling the driver-vehicle and environment, which consists of considering driver-vehicle as a rectangular hybrid input output automaton while representing contextual information about driving environment as conditions on the automaton variables. The construction of the model is ensured through a continuous monitoring of the driver-vehicle and environment system. The use of rectangular predicate states and environmental conditions will facilitate the verification of driving behavior. We then present a formal verification of properties of the constructed model expressed in Probabilistic Computational Tree Logic (PCTL) to assess its convenience to different traffic situations.

A Delay Tolerant Networks is a network where contemporaneous connectivity among all nodes does not exist. Conventional routing protocols are not suitable for DTN because they assume end-to-end connectivity. DTNs are characterized by the absence of it. This leads to the critical problem of how to route a packet from one node to another, in such a network. This problem becomes more complex, when the node mobility also is considered. In this paper, a novel strategy is presented and evaluated for controlling messages in DTNs based on the delivery predictability metric of probabilistic routing protocols. Simulation results show that this new strategy gives better results in terms of delivered and relayed messages.

Vehicular Ad-hoc Networks (VANET) are considered as a promising approach for building a variety of applications for Intelligent Transportation Systems (ITS). They are a kind of mobile networks that enables moving vehicles to exchange information about the driving environment. Although the amount of information disseminated through a VANET provides a great opportunity to enhance traffic safety, a study of the behavior of a human driver towards this information remains as an important axis to ensure safer roads. Driving behavior models are proposed by researchers as an important approach that allows a better understanding of human driving behavior. The main goal of this paper is to model and learn driver behavior in the presence of different type of traffic information. For this, we propose a new formal approach to construct a driving behavior model that will be adapted to an individual driver. To describe the model we define rectangular hybrid input output automata formalism which consists of an adaptation of a set of notions related to hybrid automata concept. Then for model construction, we propose an online passive learning based approach to construct the model according to the observed driving behavior. The constructed model may be useful to predict the driver behavior in the future, prevent unsafe situations and provide more comfort to the driver.

Testing conceptually consists of three activities: test case generation, test case execution and verdict assignment. Using online testing, test cases are generated and simultaneously executed (i.e. the complete test scenario is built during test execution). This paper presents a framework that automatically generates and executes tests "online" for conformance testing of a composite of Web services described in BPEL. The proposed framework considers unit testing and it is based on a timed modeling of BPEL specification, a distributed testing architecture and an online testing algorithm that generates, executes and assigns verdicts to every generated state in the test case.