Résumé: In today’s digital age, the rapid increase in online activities has heightened web users’ privacy risks, making them more susceptible to various cyber-attacks. Among these, phishing is one of the most widespread threats in cybersecurity. Phishing employs deceitful tactics to lure individuals and organizations into visiting malicious URLs and disclosing sensitive information, such as passwords, credit card details, and personal data. These attacks are carefully crafted to mislead users into believing they are interacting with legitimate websites or online services, to steal account information for malicious purposes. To address the limitations of traditional phishing detection methods—such as their inability to detect zero-day attacks, high rates of false positives and negatives, and the need for frequent updates to list-based approaches (e.g., blacklists and whitelists)—this paper presents an advanced approach for identifying malicious URLs by integrating transformer learning and deep learning techniques. Specifically, Distilled Bidirectional Encoder Representations from Transformers (DistilBERT) extracts features from URLs and captures relevant textual information. A hybrid deep learning model combining convolutional neural network (CNN) and long short-term memory (LSTM) layers is then applied to classify the dataset into malicious and legitimate URLs. The novelty of this method lies in combining DistilBERT with a hybrid deep learning model (CNN-LSTM) to detect malicious URLs. The proposed approach was evaluated on two large public datasets to test its effectiveness under various conditions. Experimental results demonstrate that it performs exceptionally well, achieving the highest accuracy of 98.19%, and outperforming some comparable methods in the literature on similar datasets. This proves that the integration of DistilBERT and the hybrid CNN-LSTM model is successful in detecting malicious URLs. DistilBERT optimizes the feature extraction process, while the CNN-LSTM model leverages the strengths of both models and mitigates the limitations of each, resulting in a more comprehensive approach to classification.

Résumé: In today’s digital age, the rapid increase in online activities has heightened web users’ privacy risks, making them more susceptible to various cyber-attacks. Among these, phishing is one of the most widespread threats in cybersecurity. Phishing employs deceitful tactics to lure individuals and organizations into visiting malicious URLs and disclosing sensitive information, such as passwords, credit card details, and personal data. These attacks are carefully crafted to mislead users into believing they are interacting with legitimate websites or online services, to steal account information for malicious purposes. To address the limitations of traditional phishing detection methods—such as their inability to detect zero-day attacks, high rates of false positives and negatives, and the need for frequent updates to list-based approaches (e.g., blacklists and whitelists)—this paper presents an advanced approach for identifying malicious URLs by integrating transformer learning and deep learning techniques. Specifically, Distilled Bidirectional Encoder Representations from Transformers (DistilBERT) extracts features from URLs and captures relevant textual information. A hybrid deep learning model combining convolutional neural network (CNN) and long short-term memory (LSTM) layers is then applied to classify the dataset into malicious and legitimate URLs. The novelty of this method lies in combining DistilBERT with a hybrid deep learning model (CNN-LSTM) to detect malicious URLs. The proposed approach was evaluated on two large public datasets to test its effectiveness under various conditions. Experimental results demonstrate that it performs exceptionally well, achieving the highest accuracy of 98.19%, and outperforming some comparable methods in the literature on similar datasets. This proves that the integration of DistilBERT and the hybrid CNN-LSTM model is successful in detecting malicious URLs. DistilBERT optimizes the feature extraction process, while the CNN-LSTM model leverages the strengths of both models and mitigates the limitations of each, resulting in a more comprehensive approach to classification.

Résumé: In contemporary times, the proliferation of phishing attacks presents a substantial and growing challenge to cybersecurity. This fraudulent tactic is designed to deceive unsuspecting individuals, enticing them to access malicious websites and disclose sensitive personal information such as usernames, passwords, and financial details. As a result, malevolent actors exploit this data for illicit purposes. As the sophistication and maliciousness of phishing continue to evolve, researchers are earnestly developing multiple anti-phishing solutions in the literature. Among these solutions, those based on machine learning and deep learning models have gained substantial attention in recent years. This study proposes an intelligent mechanism to detect phishing URLs. The proposed system is based on the permutation importance method to select the most relevant URL features and the SMOTE-Tomek link method to solve

Résumé: Phishing attacks pose a significant and escalating threat to cybersecurity in recent times. This deceptive scam aims to trick naive users, luring them into visiting harmful websites and sharing sensitive information, including credentials, credit card numbers, and passwords. Consequently, cybercriminals exploit this data for their own gain. As the sophistication and maliciousness of phishing continue to evolve, researchers are earnestly developing multiple anti-phishing solutions in the literature. Among these solutions, those based on deep learning models have gained substantial attention in recent years. This study proposes an intelligent, deep-learning-based mechanism to detect phishing URLs. The proposed system is based on the permutation importance method (PIM) to select the most relevant URL features, and the Smote-Tomek link method to solve the problem of an unbalanced dataset. In addition, four DL models—CNN, LSTM, and two hybrid models (CNN-LSTM and LSTM-CNN)—are tested to identify the more suitable detection model for the phishing field. The experimental results demonstrate the successful functioning of the proposed phishing detection mechanism. It is observed that the proposed mechanism achieved an accuracy ranging from 93.36% to 96.43% without feature selection and data balance across two variants of datasets and different DL classifiers. It also achieved an accuracy ranging from 94.12% to 96.88% with feature selection and data balance. Finally, our phishing detection mechanism is implemented as web application to enhance its usability for web users

Résumé: Traditional education systems do not provide students with much freedom to choose the right training of study that suits them, which leads on long-term to the negative effects not only on social, economic and mental’ well-being of student, but also will have a negative effect on the quality of the work produced by this student in the future. In addition, skills and talents in a specific area refer to general skills and working life practices. This paper presents a pedagogical orientation system based on blockchain technology and machine learning that accurately predicts the future specialized training and profession in which the common core student will have more opportunity to succeed. Python programming language has been used to implement the suggested pedagogical orienting mechanism in two key steps. (1) In the profiles matching step we have used the incidence matrix to correlate the professional and training profiles. (2) The student ranking issue was resolved at the ranking profiles step using the learning to rank methodology. The data used to feed the various machine learning algorithms in our system are stored on the blockchain, representing the academic results of 320 students at the end of the common core training in the computer science department of 20th august 1955 university in Skikda, Algeria, which served as a database for subsequent experiments and served to confirm the system's feasibility and credibility. The proposed system has demonstrated its effectiveness by accurately predicting the future specialized training compatible with the student's scientific capabilities, allowing him to succeed and excel. Also, it has contributed to reducing the possibility of falsifying results and certificates while enhancing security and transparency through blockchain protocols. Finally, this system encourages all participants to trust the automatic orientation process, which makes it possible to improve the issues experienced in the fields of education by enhancing interactions between the student, the university, and the employing industries during the educational orientation process.

Résumé: Adaptive learning has garnered researchers’ interest. The main issue within this field is how to select appropriate learning objects (LOs) based on learners’ requirements and context, and how to combine the selected LOs to form what is known as an adaptive learning path. Heuristic and metaheuristic approaches have achieved significant progress on personalized and adaptive recommendations, but the operators of some heuristic algorithms are often fixed which decreases the algorithms’ extendibility. This paper reviews existing works and proposes an innovative approach. We model the proposed approach as a constraints satisfaction problem, and an improved genetic algorithm named adaptive genetic algorithm is proposed to solve it. The proposed solution does not only reduce the search space size and increase search efficiency but also it is more explicit in finding the best composition of LOs for a specific

Résumé: Nowadays, with the variety of internet frauds, every web user while browsing the net is vulnerable to being a target of various attacks. The phishing attack is one of the largest and most effective cyber threats; it is a sort of social engineering technique employed by web hackers, with the aim of deceiving users and stealing their credentials for financial gain. The continuous growth and the rising volume of phishing websites have led researchers to propose several anti-phishing solutions to fight against this cyber-attack such as visual similarity-based approaches, list-based approaches, machine learning, heuristics-based techniques … etc, moreover deep learning in recent years has gained increasing interest in several areas, especially in the phishing detection area. In this paper, we propose a deep learning approach to detect phishing websites using convolutional neural networks testing both 1D CNN & 2D CNN

Résumé: While some studies indicate that flipped classrooms offer many positive educational outcomes, other studies draw attention to limitations associated with flipped classroom (students' limited preparation prior to class, students' need for guidance at home, students' inability to get immediate feedback while they study at home, and little research has focused on students' learning outcomes, such as: satisfaction and motivation). This paper attempts to address several of these limitations through exploratory studies conducted in an Algerian University. The approach proposed in this paper called Flipped classroom based on Context-Aware mobile learning system (FC-CAMLS) aims to provide learners with an adapted course content format based on their feedback and context. The latter has a significant influence on multimedia content in adaptive mobile learning. The system was implemented in an English Language course. It was expected that the FC-CAMLS increased the management of students' heterogeneity. A quantitative analysis by means of structural equation modeling was performed to analyze the causal relationships between knowledge, skills, motivation and students' satisfaction. The results show that the system has positive effects on students' knowledge, skills, and motivation. Finally, our research provides useful results that the use of the context dimensions and learner's feedback in adaptive mobile learning is more beneficial for learners especially in the flipped classroom.

Résumé: Students nowadays are hard to be motivated to solve logical problems with traditional teaching methods. Computers, Smartphone's, tablets and other smart devices disturb their attention. But those smart devices can be used as auxiliary tools of modern teaching methods. The flipped classroom is one such innovative method that moves the solving problems outside the classroom via technology and reinforces solving problems inside the classroom via learning activities. In this paper, the authors implement flipped classroom as an element of Internet of Things (IOT) into learning process of mathematical logic course. In the flipped classroom, an Intelligent Tutoring System (ITS) was used to help students work with the problems in the course outside the classroom. This study showed that perceived usefulness, self-efficacy, compatibility, and perceived support for enhancing social ties are important antecedents to continuance intention to use flipped classroom.

Résumé: In this paper, we show how data mining algorithms (e.g. Apriori Algorithm (AP) and Collaborative Filtering (CF)) is useful in New Social Network (NSN-AP-CF). “NSN-AP-CF” processes the clusters based on different learning styles. Next, it analyzes the habits and the interests of the users through mining the frequent episodes by the Apriori algorithm. Finally, it groups dynamically the users based on the collaborative filtering. The participants in this study consisted of 80 university students who were asked to analyze the differences in skill level when using various learning activities. Moreover, 40 students were included in this study in order to examine the effectiveness of NSN-AP-CF. The experiment results proved that the proposed algorithm, which considers the grouping dynamically the users and the discovery of all frequent episodes, generates better precisions compared with the other algorithms (F1 = 0.649)

Résumé: A large number of studies attest that learning is facilitated if the teaching strategies are in accordance with the students learning styles (LS), making the learning process more effective and considerably improving student's performances. But, traditional approaches for detection of LS are inefficient. This work determines the current preferences through dynamic Bayesian network that represent the matches between LS and teaching strategies in order to determine how much a given strategy is interesting to a student. The LS theory that supports this approach is the LS model proposed by Felder-Silverman's learning styles model (FSLSM). Our approach gradually and constantly adjusts the student model, taking into account students' performances, student's effort, student's intensity, student's resistance and student's attention. Promising results were obtained from experiments, and some of them are discussed in this paper.

Résumé: Adaptive E-learning Hypermedia Systems (AEHS) are an innovative approach to a web learning experience delivery. They try to solve main shortcomings of classical hypermedia e-learning applications: “one-size-fits-all” approach and “lost-in-hyperspace” phenomena by adapting the learning content and its presentation to needs, goals, thinking styles and learning styles of every individual learner. This paper outlines a new approach to automatically detect learner's thinking and learning styles, and taking into account that thinking and learning styles may change during the learning process in an unexpected and unpredictable way. The authors' approach is based on the Felder Learning Styles Model and Hermann thinking styles model.

Résumé: An Adaptive E-learning Hypermedia System (AEHS) is an approach whose target is to personalize the learner's learning experience. A number of AEHS have been developed to support thinking styles as a source for adaptation. However, these systems suffer from several problems, namely: lack of maintenance adaptation to thinking style and the insertion of specific teaching strategies into learning content. This paper proposes an AEHS model based on thinking styles and knowledge level. On one hand, the developed prototype will assist a learner in accessing and using learning resources which are adapted according to his/her personal characteristics (in this case his/her thinking style and level of knowledge). On the other hand, it will facilitate the learning content teacher in the creation of appropriate learning objects and applying them to a suitable pedagogical strategy. Keywords: Adaptive hypermedia, thinking style, ontology, intelligent system, learner model

Résumé: Recent years have shown increasing awareness for the importance of adaptivity in e-learning. Since the learning style of each learner is different. Adaptive e-learning hypermedia system (AEHS) must fit different learner’s needs. A number of AEHS have been developed to support learning styles as a source for adaptation. However, these systems suffer from several problems, namely: lack of maintenance, adaptation to learning style, less attention was paid to thinking styles and the insertion of specific teaching strategies into learning content. This paper proposes an AEHS model based on thinking styles and knowledge level. On one hand, the developed prototype will assist a learner in accessing and using learning resources which are adapted according to his/her personal characteristics (in this case his/her thinking style and level of knowledge). On the other hand, it will facilitate the learning content teacher in the creation of appropriate learning objects and their application to suitable pedagogical strategies.

Résumé: A distinct feature of an adaptive e-learning hypermedia system (AEHS) is the learner model it employs, that is, a representation of information about an individual learner. Learner modeling and adaptation are strongly correlated, in the sense that the amount and nature of the information represented in the learner model depend largely on the kind of adaptation effect that the system has to deliver. In fact, we see a problem arising when teachers assume similar learning styles, thinking styles, levels of knowledge and abilities for learners. This is because learners that are less able will feel that it is too difficult for them to follow and those that are more capable will feel as though the learning method is too easy. Teachers can adjust the standards; however, there may be conflicts between learners with varied styles. Thinking style, learning style, level of knowledge, preferences and ability of learner are part of learner’s characteristics, which have significant influence on the activity of learners in the learning process. In this paper we have focused our attention on the learner model, which allows for the discovery of preferences, needs and interests of users that have access to an AEHS. In order to observe the psychological and pedagogical characteristics of a learner, a quantitative and qualitative research is conducted based on a questionnaire. The thinking style of a learner is analyzed. Based on the statistical results, we figure out the rules about pedagogical activity decision-making. This study presents two subsequent experiments. The first experiment explores the relationship of thinking style and pedagogical activities to validate this specific psychological construct in the context of an AEHS. The second experiment reduces the questionnaire to 60 questions, using a filtering principle keeping the validity of the original questionnaire.

Résumé: A number of adaptive e-learning hypermedia systems (AEHS) have been developed to support learning styles as a source for adaptation. However, these systems suffer from several problems, namely: lack of maintenance adaptation to learning style, less attention was paid to thinking styles and the insertion of specific teaching strategies into learning content. This paper proposes an AEHS model based on thinking styles and domain ontology. The experiment was completed in three phases for both experimental and control groups. In the first phase all the students were informed that they will participate in an experimental process. The students received a short introduction on how to use the system and to create a user account for login purposes into the system. Then, information about thinking styles categories were given to the experimental group and were asked to complete the questionnaire. In the second phase, the students followed regularly the lessons until the completion of the course; meanwhile taking a quiz at the end of each lesson. In the third phase, learners followed a link to do the post-test.

Résumé: The approach proposed in this paper called Flipped classroom based on Context-Aware mobile learning system (FC-CAMLS ) aims to provide learners with an adapted course content format based on their feedback and context. The latter has a significant influence on multimedia content in adaptive mobile learning. The contribution was applied in the context of the flipped learning in order to manage the heterogeneity of context imposed by this approach. Firstly, we present a quantitative analysis by means of structural equation modeling to analyze the causal relationships of knowledge, skills, and motivation with students' satisfaction. Secondly, we confirm that the proposed flipped classroom has positive effects on students' knowledge, skills, and motivation. Finally, our research provides useful results that the use of the context dimensions and learner’s feedback in adaptive mobile learning is more beneficial for learners especially in the flipped classroom.

Résumé: Flipped classroom can be used to encourage teachers to prepare their own stories for their students and connect with peers to build their own collaborative learning spaces. Teachers can create digital storytelling from the content or have their students do it to demonstrate their understanding of the content. The greatest benefit in the flipped classroom may be found when students may be given assignments in which they are asked to research a topic, look for pictures, record their voice and then choose a particular point of view. This paper explores storytellers’ experiences of digital storytelling (DST) through a flipped classroom approach. A mixed research method was employed, using multiple sources of data collection, including pre-and post-tests , Perception of Flipped Learning Experience questionnaire, the teachers’ in-class observations, and semi-structured focus-group interviews. The results revealed that the flipped classroom not only enhanced the participants’ motivation, making them more active, but also significantly improved their ability.

Résumé: Learners usually meet cognitive overload and disorientation problems when using e-learning system. At present, most of the studies in e-learning either concentrate on the technological aspect or focus on adapting learner's interests or browsing behaviors, while, learner's skill level and learners' success rate is usually neglected. In this paper, the authors propose an online course generation based not only on the difficulty level of a learning unit, but also the changing learning performance of the individual learner during the learning process. Therefore, considering learner's skill level and learners' success rate can promote personalized learning performance. Learners' skill level is obtained from pre-test result analysis, while learners' success rate is acquired through specific tests after completing a learning unit. After computing success rate of a learning unit, the system then modifies the difficulty level of the corresponding learning unit to update courseware material sequencing. Experiment results indicate that applying the proposed intelligent e-learning system can generate high quality learning paths, and help learners to learn more effectively.

Résumé: Blockchain is a shared database that builds trust without third-party involvement. Its decentralized architecture ensures data is dispersed and cannot be wiped. This makes blockchain highly suitable for e-learning and online education, which have become primary means for obtaining certifications and degrees, especially with the surge in online learning due to the COVID-19 pandemic. The increase in online universities and platforms has led to issues with fraudulent coursework, poor evaluations, and degree verification. This project proposes a blockchain-based strategy for the education sector, using smart contracts to manage student data and provide success certifications

Résumé: One of the toughest cybersecurity issues today are insider threats for which no good solution has been provided to limit their damage through commonly used security solutions. Which could cause serious damage to the assets of the organization. This article begins by presenting a comprehensive multidisciplinary survey of insider threats and the various AI technologies in use. Then, we proposed a method to detect internal threats by analyzing user behavior. A series of activities and events that are analyzed to extract features to effectively detect malicious activity from normal activity. The selected feature vectors are used to train the model. The convolutional neural network (CNN) is used during the implementation phase to detect insider threats of fixed-size feature vectors. The data used to conduct the experiments is public data related to internal threats cert r4.2. The experimental results of the proposed model

Résumé: The insider threat has captured the attention of a large number of researchers, as a sensitive and critical issue for most organizations in today’s digital world. It is also a major source of information security and can cause more damage and financial loss than any other threat. In this article, we’ve used feature engineering for features that represent users’ day-to-day activities. We tried different machine learning models such as random forest, xgboost and Catboost. Since the data used to detect malicious activity is unbalanced, the target audience is small. We used KMeansSmote to balance the classes of learning so that the algorithms can learn both classes well. And we used the catboost algorithm to identify the malicious user. The dataset used to evaluate this model is Cert v4.2. CatBoost outperformed other models with the highest F1-score of 95%

Résumé: The internal and external security of a company is important. External security can be secured by setting up mechanisms to monitor any external flow, while internal security is the most complex, in this case how do we monitor internal workers who have full privileges to access the resources and data of the organization? All necessary measures must be in place to avoid internal damage, which has increased considerably in last years. Since the number of harmful behaviors is very low compared to normal events, the imbalance in class scores does not allow supervised learning algorithms to provide accurate results as their learning depends on balanced categories. Therefore, it is necessary to use a model capable of distinguishing clearly the harmful category. In previous work, ML techniques were used, although they are less effective if the data used are not balanced. In this document, we propose an S-LSTM

Résumé: Adaptive and context-aware learning provides learning content according to a learner's context. In order to achieve this goal, the dimensions that constitute the context in the current learner's state have to be determined. There are several existing works within this field and each of these are taking care of a subset of context parameters - like learning styles, learner location, etc. But, a standardized model that helps to capture a learner's context in its entirety is not available. The requirement to define context more precisely and in a uniform way has been identified by several researchers because a general and precise definition of context can facilitate the identification of what does and does not constitute context and can enable reuse and share of contextual data over applications. To this end, this work proposes a reference context model that helps to define a learner's context. The proposed model is developed by consolidating the various context parameters used in the existing adaptive systems and organizing them into an appropriate structure.