Program

S1: ML Applications

onsite

S2: Computer Vision

onsite

S3: IoT

online

S4: Big Data

onsite

S5: ML Applications

Hybrid

S6: IoT

Hybrid

S7: Computer Vision

online

S8: GL

Hybrid

S9: Cybersecurity

onsite

S10: Healthcare

onsite

S11: Machine Learning

onsite

S12: Cloud Computing

onsite

S13: Cybersecurity

onsite

S14: Healthcare

onsite

S15: NLP

onsite

S16: Quantum Computing

Hybrid

S1.1: Powdery mildew disease classification in laboratory and real-field images using convolutional neural networks for precision agriculture.

Dounia Kawther Dihya Bourzig, Mansour Abed, Mostefa Merah ( Abdelhamid Ibn Badis University– Mostaganem ) 

Precision agriculture encounters challenges in detecting plant diseases. Powdery mildew is a fungal disease that impacts various crops. Accurately identifying this disease in agricultural field images is crucial for effective crop management. This research presents a systematic approach to assess the generalization capabilities of convolutional neural network (CNN) models, specifically DenseNet121, EfficientNetV2B0, InceptionResNetV2, InceptionV3, MobileNet, and ResNet50V2, in classifying powdery mildew disease when transitioning from laboratory images to real-field images. The experiment is organized into three steps, utilizing distinct subsets derived from the PlantVillage and Kaggle plant disease recognition datasets. PlantVillage supplies laboratory images, while the plant disease recognition dataset provides real-field images. The data is divided into five subsets, allowing for a comprehensive analysis across six training and evaluation scenarios. The results indicate that while all models achieve perfect accuracy in controlled laboratory conditions, their performance significantly declines when assessed on real-field data, underscoring the challenges of generalizing to complex, real-world environments. However, models trained on real-field data, or a combination of both datasets, demonstrate strong performance in real-field applications, with some CNN models achieving perfect accuracy. InceptionResNetV2 and InceptionV3 consistently excelled across various scenarios, positioning them as promising candidates for smart farming applications aimed at accurately classifying powdery mildew disease. This approach holds substantial potential for the agriculture industry, establishing a solid foundation for future studies.

S1.2: Leveraging Deep Learning to Automate the Approval Process for Durum Wheat Grain Quality

Ghalia Merzougui*, Nadia Boulelouah**, Abdelheq Mokhtari*, Selsabil Maalem*, Saadi Leila*, Abderraouf Hebira**, Imene Saadoune*, Hanine Merzougui*,  Aya Koulib*   (* University of Batna 2,  Algeria , ** University of Batna 1, Algeria)

This study addresses the urgent need for accurate and efficient quality control of durum wheat grains in Algeria, where traditional inspection methods are often insufficient. To tackle this, we trained and evaluated seven convolutional neural network (CNN) models on a comprehensive dataset of 18,362 images, aiming to classify wheat species, varieties, and impurities. Our findings demonstrate the ResNet50 model's superior performance in species and variety classification, achieving accuracies of 99.59% and 98.11%, respectively.
Although its accuracy for impurity detection was comparatively lower, likely due to the limited size of the impurity’s dataset.
ResNet50 still demonstrated its potential in addressing this challenging classification task. Future directions for this study include integrating the ResNet50 model into an automated quality control system, representing a significant step forward in enhancing the efficiency, accuracy, and consistency of wheat quality assessments within the Algerian agricultural sector. This adaptable solution could extend to other cereal types, offering a scalable impact on agricultural quality control practices more broadly.

S1.3 Deep Learning-Based Detection of Apple Leaf Diseases Using Image Analysis

Bedda Hana, Saadna Yassmina (University of Batna 2,  Algeria)

In the dynamic field of agriculture, the intelligent diagnosis and classification of plant diseases represent a key research objective. Among widely cultivated fruits, apples are prominent, and diseases such as "Scab" and "Rust" are prevalent in apple plants. Differentiating between these diseases, as well as identifying "multiple diseases" and healthy apple leaves, poses a challenging task. To address this issue, this research proposes an advanced deep learning algorithm capable of accurately classifying and recognizing various apple leaf diseases with minimal loss. Utilizing the 'Plant Pathology 2020 - FGVC7' dataset with 1821 images categorized into Scab, Rust, Healthy, and Multiple diseases, our study embarks on an in-depth analysis of apple leaf conditions. Furthermore, the issue of imbalanced datasets is mitigated by implementing the SMOTE algorithm. To determine the optimal model, multiple approaches are explored, including modifying learning rate values, employing different batch sizes, and utilizing diverse optimizer functions. The study encompasses the application of both a simple CNN model and transfer learning models, such as DenseNet, InceptionV3, VGG16, ResNet50, and ResNet101. Through extensive experimentation, it is revealed that the DenseNet model outperforms the others, achieving a remarkable validation accuracy of 97.18% with a learning rate of 0.0001, a batch size of 64, and the "Adam" optimizer function. The findings demonstrate the potential of deep learning algorithms for revolutionizing apple disease detection, offering farmers and agriculturalists a reliable tool to combat plant diseases effectively.

S2.1: Local ternary descriptors and CNN fusion for facial expression recognition.

Chebah Ouafa, chebah anissa (University of Annaba)  

In this paper, we present a multimodal classification approach based on a local ternary descriptors and a combination of Convolutional Neural Network (CNN) to recognize facial expressions. Initially, we need to perform the pre-process of face image, which includes face detection, face image cropping and local ternary descriptors (LTP and GLTP) calculation. Next, we construct a cascade CNN architecture using the multimodal data of each descriptor (CNNPLTP, CNNNLTP for LTP operator and CNNPGLTP, CNNNGLTP for GLTP operator) to extract facial features and classify facial images. Finally, we apply separate classifiers for each modality and then we combine the four multimodal outputs using a product method to obtain the final decision of our system. Experimental results using CK+ and JAFFE datasets show that the proposed multimodal approach achieves superior recognition performance compared to the existing studies with classification accuracy of 98. 60 % and 95. 45 %, respectively.

S2.2: Enhancing Image Classification with Ensemble Deep Learning through Deep Feature Concatenation

Boudouh Nouara, Bilal Mokhtari (University of Biskra)

Ensemble deep learning represents a powerful methodology that combines multiple deep learning models to improve overall performance. By aggregating predictions or features from these models, ensemble methods achieve superior accuracy, robustness, and generalization compared to individual models. In this paper, we introduce a novel approach to enhancing classification performance through ensemble deep learning techniques. Our method leverages the final extracted features from several pre-trained convolutional neural networks (CNNs). Each CNN independently generates a feature vector, and these vectors are concatenated to integrate the diverse outputs of each model. The resulting concatenated feature vector is subsequently fed into a classifier, specifically fully connected layers, which enhances the robustness and generalization capabilities of the classification task. This approach demonstrates that concatenating final extracted features from different models can significantly elevate classification accuracy. Experimental results on the Cats vs Dogs dataset reveal notable improvements in classification accuracy over single-model approaches. Specifically, VGG16 and VGG19 achieved accuracies of 91.71\% and 90.18\%, respectively. The ensemble of outputs from VGG16 and VGG19 resulted in a substantial accuracy boost, reaching 93.75\%. Inception-V3 exhibited a higher accuracy of 97.75\%, and when combined with VGG19 and VGG16, it achieved accuracies of 97.96\% and 98.12\%, respectively. The ensemble comprising Inception-V3, VGG16, and VGG19 yielded the highest performance, with an accuracy of 98.48\% and a low error rate of 4.50\%. Furthermore, EfficientNetB0 recorded an accuracy of 96.67\%, which improved to 98.32\% when integrated with Inception-V3. Notably, MobileNet achieved an accuracy of 98.00\% with an error rate of 0.09\%, while the combination of MobileNet and VGG16 further enhanced performance, resulting in an accuracy of 99.00\% and an error rate of 0.18\%.

S2.3 Malware Classification on Malimg Using MobileNet and LSTM for Efficient Detection

Hayet Benbrahim, Ali BEHLOUL  (University of Batna2)

Malware detection is essential for protecting systems from evolving threats, with traditional methods struggling to keep up with increasingly sophisticated malware. In this study, we present a hybrid model combining MobileNet's efficiency with LSTM's sequential learning to enhance malware classification. Trained on the Malimg dataset, the model achieved a high test accuracy of 99.26%, surpassing existing approaches. The results demonstrate its effectiveness and suitability for deployment in resource-constrained environments, offering a promising solution for real-time malware detection.

S3.1: Machine Learning and Deep Learning for Enhanced DDoS Detection in IoT-based Intrusion Detection System.

Kamir Kharoubi, Sarra Cherbal, Maroua Akkal (university of Ferhat Abbas Setif 1 )  

Securing Internet of Things (IoT) networks is critical due to their increasing prevalence and susceptibility to cyberattacks. Machine Learning (ML) and Deep Learning (DL)-powered Intrusion Detection Systems (IDS) provide a promising solution to counter these threats. This study focuses on detecting Distributed Denial of Service (DDoS) attacks, ensuring IoT device availability and protecting critical services within IoT ecosystems. Five ML and DL models, including XGBoost, were trained and evaluated on the recent CICIoT2023 dataset to classify network traffic as benign or DDoS. Evaluation results show that XGBoost outperformed other models, achieving a 99.96\% accuracy in multi-classifying DDoS attack types with benign traffic and a rapid prediction time of 0.56 microseconds per traffic sample. These results highlight XGBoost's potential for real-time IoT security applications.

S3.2: Enhanced DCOPA with Security Features for WSN-based IoT Clustering

Foudil Mir (University of Bejaia), Zoubeyr Farah (University of Bejaia) and Farid Meziane (University of Derby, Derby, UK)

In the context of Wireless Sensor Networks (WSN) applied to the Internet of Things (IoT), security is a major concern due to the vulnerability of these networks to various threats and attacks. While the DCOPA protocol is efficient in terms of energy management and clustering, its security aspects have not been addressed. To mitigate security risks such as Sybil attacks, identity spoofing, and replay attacks, this paper proposes an enhancement to the DCOPA protocol, called Secured DCOPA (SECDCOPA), by integrating a robust and lightweight security mechanism into its operation.The main objective is to enhance the security of the network constructed using DCOPA to mitigate security threats while preserving energy efficiency. The proposed mechanism ensures that only authenticated entities can join the established cluster network, guaranteeing the authenticity of communicating entities and providing complete control over the integrity of exchanged information. The integration of this security mechanism into DCOPA is of paramount importance in IoT environments, where security is under constant threat.By employing a distributed hash-based security approach, we enhance the security features of DCOPA while preserving its energy efficiency. This ensures the network's resilience against attacks and maintains its stability.

S3.3 A Data Communication based on Deep Learning Model for Efficient IoT Energy Clustering

Foudil Mir (University of Bejaia), Dalil Hadjout (Sonelgaz Distribution, Bejaia, Algeria) , Abderazzak Sebaa (Higher School of Computing and Digital Sciences and Technologies ESTIN, Amizour, Bejaia), Abdelkader Laouid (University of El Oued, Algeria) and Farid Meziane (University of Derby, Derby, UK)

The emergence of artificial intelligence, especially, deep learning combined with time series forecasting has attracted increasing attention from academics and researchers in various domains. This paper proposes a new energy-efficient clustering protocol for data communication based on the DCOPA protocol using a deep learning model based on a Multilayer Gated Recurrent Unit (GRU) with a Bayesian optimizer. The core idea is to train the network nodes to reproduce their behaviors during the execution of the DCOPA protocol and create a deep learning model based on the training with the real values of the CHs election rounds generated during many executions of the DCOPA protocol. Hereafter, the deep learning model will be deployed to forecast a future set of CH nodes in each round for the future rounds. Therefore, the large volumes of exchanged data during the clusters' formation will no longer be required. To validate the proposed solution, we conducted a large number of experiments with the DCOPA protocol and recorded the results in a history log. The obtained results are very encouraging to continue this work.

S3.4: Enhancing Online Petitions in Somalia for Civic Engagement and Policy-Making Using Blockchain

Mohamed Ibrahim Abdullahi, Abdirahman Abdiaziz Geesey, Abdirahman Ali Maow, Mohamed Hasan Omar, Bashir Abdinur Ahmed (Jamhuriya University of Science and Technology Mogadishu, Somalia)

To achieve a stable and equitable government in Somalia, it is crucial to encourage active public participation in politics and informed decision-making. An online petition is a digital form that allows people to gather support and signatures for a cause or request. Traditional methods of civic engagement often face significant challenges, such as limited access, safety concerns, and the inherent weaknesses of centralized systems, which are prone to issues like manipulation and fraud. Although citizen involvement is crucial for influencing public policy, creating a secure and transparent online petition system specifically tailored to Somali issues remains a significant challenge. This study explores the potential of blockchain technology to enhance online political campaigns and address these issues by developing a decentralized solution. Unlike other platforms that adopt a global or generalized approach, the proposed system is designed to meet the unique needs of the Somali context. It ensures the integrity of data and the authenticity of petition signatures, thereby fostering trust in the process. The platform features a user-friendly interface that enables the creation, signing, and management of petitions, along with real-time monitoring and interactive tools to promote community involvement. Additionally, the petition data provides policymakers with valuable insights, facilitating decisions that align with the community's needs. In conclusion, this research aims to advance community engagement and policy-making in Somalia by developing a decentralized, transparent, and secure blockchain-based online petition platform.

S4.1: Big Data Analytics for earthquake management: A review

Khedoudja Bouafia, Hachem Slimani (University of Bejaia), Hassina Nacer (University of Science and Technology Houari Boumediene, Algiers, Algeria), Hamida Seba (LIRIS Laboratory, Claude Bernard University, Lyon 1, Lyon, France)  

Earthquakes are unexpected geological events that happen in some seconds, and they cause enormous damage in terms of both human and material losses. That is why earthquake and seismic risk preparedness is essential and vital to reduce the negative impact of these events. In this framework, Big Data (BD) technologies offer new possibilities for improving understanding, prediction and response to earthquakes. Furthermore, Big Data Analytics (BDA) offers numerous possibilities for improving emergency situations particularly earthquakes, whose collected data can be rapidly analyzed to identify areas at risk, monitor the evolution of the situation and guide interventions. In this paper, we present a review of the main state-of-the-art approaches based on BDA techniques for earthquake management. In this setting, we propose a new classification that allows to structure the main existing works, dedicated to the studied subject, into relevant categories. This classification has allowed us to understand especially the difficulties to set up in practice the theoretical results and conclusions of some works; and moreover the importance to investigate more the issue by involving advanced BDA techniques combining with new technologies in particular Mobile Internet of Things (MIoT). To achieve this, we have highlighted some relevant open issues and research directions that can be worth to be investigated as future works.
Index Terms : Earthquake management, Big Data Analytics, Cutting-edge technologies.

S4.2: Benchmarking Service Recommendations Made Easy with SerRec-Validator

Ayoub Mokeddem, Saber Benharzallah, Chafik Arar, Tahar Dilekh (University of Batna 2), Laid Kahloul (University of Biskra), Hamouma Moumen (University of Batna 2)

Service recommendation is a research field that has gained significant attention in recent years due to its potential to enhance the efficiency and effectiveness of service provision. However, one prominent weakness in this field is the evaluation of service recommendation models, which plays a crucial role in assessing their performance and effectiveness. While various evaluation metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Square Error (MSE) are commonly used in research studies, these metrics often fail to reflect the real-world performance of the system, making the evaluation of recommender systems a challenging problem that requires careful consideration and research.

This paper aims to address the challenges associated with benchmarking and evaluating service recommendations by presenting a new framework called ”SerRec-Validator”. The framework provides a comprehensive set of evaluation metrics and use cases to facilitate benchmarking and evaluation processes.

S4.3 Game-Theoretic Approaches to Data Clustering - A survey

Dalila Kessira (University of Bejaia), Mohand Kechadi (University College Dublin) 

Data clustering is a widely researched discipline, yet lacking a unique solution for all clustering problems has stimulated many research efforts using several techniques. This study offers a survey of game-theoretic approaches to data clustering, providing an overview of the critical methods using cooperative and non-cooperative games in the data clustering problem. Our work primarily aims to assist researchers by compiling essential work in the field while giving some insights about the nature of the research and achieved results. We explore various models, their applications, and how they compare to traditional clustering techniques. By analyzing the strengths and limitations of these approaches, we provide a good understanding of game theory's contribution to advancements in data clustering.

P1: Industrial internet of things for smart factories : Applications, Challenges and New trends.

Tayeb AOUFI, Karima AKSA (University of Batna 2, Algeria)

P2: Mitigating DoS Attacks in IoT systems with Blockchain: A Literature Review

Chahrazad Adouane, Sonia Sabrina Bendib, Moumen Hamouma (University Batna 2, Algeria)

P3: Enhancing Deep Learning Performance with Data Preprocessing and GAN-Based Augmentation

Zouleikha Betaitia*, Aida Chefrour**, Samia Drissi*, Soufiane Khedairia* (* Mohamed Cherif Messaadia University
Souk Ahras, ** Badji Mokhtar University Annaba)

P4: Harnessing Optimization Techniques for Enhanced Performance in Elastic Optical Networks: RSA and RMSA Perspectives

Imene Achouri, Malika Babes (Badji Mokhtar Annaba University), Youssouf Achouri (University of Batna 2, Algeria)

P5: Generative Adversarial Networks in Recommendation Systems: Opportunities and Challenges

Nadhir Cheriet, Abdelouahab Belazoui (University of Batna 2)

P6: Enhancing Arabic News Recommendation System with Self-attention Mechanism

Karboua Sabrina, Fouzi Harrag (Ferhat ABBAS University Setif 1, Algeria)

P7: Blockchain Technology for Organic Product Traceability: Methods, Innovations, and Challenges in Supply Chains

Lina Moumni, Hamouma Moumen, Lahcene Guezouli (University of Batna 2, Algeria)

S5.1: An Accurate Iris Recognition System Based on Deep Transfer Learning and Data Augmentation

Abdessalam Hattab, Ali BEHLOUL (University of Batna 2), Wassila Hattab (university of Biskra) 

Iris recognition is a highly secure biometric identification method that utilizes the distinct and intricate patterns of the iris, which remain consistent throughout a person's lifetime. Remarkably, it can even distinguish between identical twins, highlighting its precision and reliability. However, despite these advantages, challenges such as reflections, occlusions, varying lighting conditions, blur, and other environmental factors can significantly degrade the performance of iris recognition systems. Deep Convolutional Neural Networks (CNNs), known for their success in image recognition, often require numerous parameters, increasing computational time and resource demands. To address these challenges, we propose an advanced iris recognition system based on Deep CNNs. Our system utilizes YOLOv4-tiny for accurate iris region detection, and for the recognition task, we introduce a deep CNNs model inspired by the pre-trained Xception, which contains fewer than three million parameters. To effectively train this deep CNN on small iris datasets, we apply Transfer learning and Data Augmentation techniques, which improve accuracy while reducing the risk of overfitting. Through two-fold cross-validation, our system achieves impressive accuracy rates of 99.81% and 99.64% on the CASIA-Iris-Interval and IITD datasets, respectively. Our method demonstrates a clear advantage over existing methods in terms of accuracy and reliability, making it a promising solution for various applications such as border control, law enforcement, and access control.

S5.2: An Enhanced Bat Algorithm Using Clustering Coefficient for Community Detection in Complex Networks

Salaheddine TAIBI, Salim BOUAMAMA, and Lyazid TOUMI (University Setif 1 - Ferhat Abbas) 

The discrete bat algorithm (DBAT) has gained significant interest in solving a variety of combinatorial optimization problems. Discovering community structures in complex networks is particularly challenging, as it is an NP-hard problem.
To address this, various adaptations of the discrete bat algorithm have been developed. This work proposes an improved discrete bat algorithm based on the clustering coefficient, called CCDBAT, to discover community structures in complex networks. The proposed algorithm generates an initial population of solutions using the clustering coefficient metric, followed by a fast local move technique to enhance the intensification process. Additionally, a random neighbor communities technique is employed to generate a random solution that guides the search process toward the global optimum. Experiments conducted on synthetic and real-world networks demonstrate that the proposed approach outperforms existing bat algorithms and exhibits improved convergence rates.

S5.3: Colouring LiDAR point cloud of wall paintings from images based on ICP

David Jesenko (University of Maribor, Slovenia),Aljaž Žel ( University of Maribor, Slovenia), Andraž Krivic ( IGEA d.o.o, Slovenia ), Andreja Pondelak ( Slovenian National Building and Civil Engineering Institute, Slovenia), Sabina Dolenec ( Slovenian National Building and Civil Engineering Institute, University of Ljubljana, Slovenia), Ana Brunčič ( Slovenian National Building and Civil Engineering Institute, Slovenia), Marko Bizjak ( University of Maribor, Slovenia) 

The sustainability of cultural heritage is one of the grand challenges of the contemporary world. Part of preservation is long-term monitoring, and LiDAR techniques are exceptionally useful. Despite all its advantages LiDAR is still not perfect, as individual points do not have a specific colour. In this article, we present a procedure for colouring a LiDAR point cloud of the wall painting from images based on the iterative closest point algorithm. As the results show, the presented methodology is suitable for colouring the LiDAR point cloud of wall paintings.

S6.1: Industrial internet of things for smart factories : Applications, Challenges and New trends.

Tayeb AOUFI, AKSA Karima (University of Batna 2) 

The Industrial Internet of Things (IIoT) is revolutionizing the manufacturing sector by transforming traditional factories into smart factories. This paper investigates the diverse applications, challenges, and current trends of IIoT in smart manufacturing. IIoT enables enhanced connectivity, sustainability, operational excellence, and quality product. In this paper, we clarify 10 applications and 10 challenges for IIoT in smart manufacturing, the key applications include predictive maintenance, Energy Management Real-Time Monitoring and Control, Data Analytics and Insights, and Supply Chain Optimization. However, the integration of IIoT also presents several challenges, such as data privacy and compliance, costs and return on investment, Skill and Knowledge Gaps and the Legacy System Integration. We highlight new trends influencing the future of smart manufacturing through IIoT technology, including the interaction in industry 5.0, the adoption of edge computing, the advent of 6G networks, and the development of AI-algorithm in order to enhance decision-making for manufacturing operations. This paper provides a detailed overview of IIoT in smart factories, addressing applications and challenges, as well as identifying future research and development paths that will motivate researchers to further investigate the topic.

S6.2: Advanced Sensors Network Design for Real-Time Livestock Watering Point Monitoring

Fatou Diop, Babacar Mbaye Faye, and Ibrahima Niang (Cheikh Anta Diop University Dakar, Senegal)

Efficient management of water resources is essential for sustainable livestock farming. Contamination of water sources by pathogens, chemicals or sediments can lead to serious health problems in livestock and affect growth rates, reproduction and milk production, impacting on economic results. Recently, innovative technologies such as the Internet of Things and cloud computing have improved water management and monitoring in various agricultural contexts. However, most proposed solutions often face limitations in real-time data processing and latency issues, particularly in remote agricultural areas with limited internet connectivity. These challenges hinder immediate response times and accurate real-time monitoring essential for livestock watering point management.
In this paper, we propose an innovative and advanced sensor network design leveraging Internet of Things (IoT), Edge-AI technologies to enhance real-time data collection and processing at remote watering points. By reducing latency and improving reliability, the proposed architecture aims to overcome traditional aims to overcome current limitations in traditional approaches. The system integrates advanced sensor/biosensors technologies with local processing capabilities to achieve timely detection and response to water availability and quality issues. Case studies demonstrate the feasibility and effectiveness of the proposed solution in improving livestock management practices.

S7.1: U-SPIN PYRAMID: U-NET WITH SPATIAL AND INTERACTION SPACE GRAPH REASONING FOR IMPROVING ROAD EXTRACTION FROM SATELLITE IMAGES

Belaid Mohammed BELAID, Lila MEDDEBER, Mohamed IMAM, Tarik ZOUAGUI ( University of Science and Technology of Oran-Mohamed Boudiaf (USTOMB)) 

Remote sensing images are essential nowadays as they allow us to extract relevant information intended for various applications, notably road networks. Today, road extraction is crucial for updating maps, urban planning, developing autonomous driving and even quantifying the damage caused to roads in the event of natural disasters. However, this task presents many challenges, including the complex geometric paths of roads, occlusion by other objects like trees and buildings, and similarities to other objects such as car parks and house surfaces. All of this makes extracting roads using only simple Convolutional Neural Networks (CNNs) difficult. Our study focuses on how the Spatial and Interaction Space Graph Reasoning (SPIN) module can improve the performance of road extraction. In this article, we chose the U-Net model to create three variants of the SPIN pyramid called U-SPIN. The three architectures demonstrate superiority over the U-Net model, outperforming the U-Net model in terms of accuracy, recall, F1 score, and Jaccard Index (IoU) on the Deepglob dataset. 
Index Terms—Remote sensing, Road network, SPIN pyramid, U-net , deep learning, CNN.

S7.2: Efficient Multi-Head Attention for Human-Object Interaction Recognition and Video Scene Graph Generation

Anfel AMIRAT, Nadia Baha, Mohamed El Amine BENTARZI, Zoheir BESSAI ( USTHB ) 

Comprehending human-object interactions in video sequences is one of the fundamental problems in visual classification and an essential step toward more detailed scene understanding. Scene graph generation (SGG) has emerged as a powerful tool to model these relationships by representing interaction as graph structures. However, while capturing temporal relationships, dynamic video scene graph generation methods often introduce significant computational complexity and struggle with long-tail distributions, making rare interactions challenging to detect. This paper proposes a static approach for scene graph genenration that leverages multi-head attention mechanisms. Each attention head is specialized in different type of relationships, such as spatial, contact, and attention-based interactions. The proposed method offers a more efficient and accurate way to recognize multi-type relationships by refining visual features through contextual information and focusing on the most relevant aspects of each interaction. We conclude this paper by comparing the proposed method with the static and dynamic state-of-the-art approaches for video scene graph generation and demonstrate the effectiveness of our method over the Action Genome dataset.

S7.3 High Performance UORA Protocol Designed for IoT-based Healthcare Applications Running on the EHT WLANs.

Said OULD AMARA and Mohand YAZID (University of Bejaia)

The rapid expansion of Internet of Things (IoT) has revealed new challenges in various domains, especially in healthcare. IoT-based healthcare applications provide the capability to upgrade patient monitoring, diagnostics and treatment by enabling real-time data collection and analysis. To meet these challenges, the following requirements must first be guaranteed by the communication interface used, namely: wide bandwidth, high throughput, low latency, and reliable data transmission. The forthcoming IEEE 802.11be WLANs (Wireless Local Area Networks) standard introduces a range of innovative features and advanced techniques in order to meet the requirements of IoT applications. One of the prominent transmission techniques ensuring real-time communications is called OFDMA (Orthogonal Frequency Division Multiple Access), it is initially introduced in the IEEE 802.11ax standard. To manage the OFDMA communications in UL (Up-Link) direction (from devices to access point), UORA (Up-link OFDMA Random Access) is the only existing standard MAC protocol. In this paper, we aim at proposing and assessing a new version of UORA protocol called HP-UORA (High Perfoamnce UORA) whose basic principle is reducing waiting delays and effeciently allocating radio resources. The obtained simulations results demonstrate the effectiveness of the proposed protocol.

S8.1: Applied Gaming-Based Emotion-Driven on Disaster Resilience Training

Hayette hadjar (Hagen University , Germany), Matthias Hemmje ( Hagen University , Germany ), Hadjadj Zineb ( CERIST, Algeria ), Binh Vu ( SRH University Heidelberg, Germany ), Abdelkrim MEZIANE (CERIST) 

Managing stress in disaster response environments is a critical challenge that requires effective strategies to enhance the resilience and well-being of emergency responders. This study introduces DisasterPlay, a prototype web-based platform designed for resilience training. The prototype features a comprehensive model design, user interface, and implementation using WebXR, facial emotion monitoring, and contactless vital signs monitoring. This approach not only improves the training experience but also aids decision-makers in selecting the most suitable candidates for high-stakes tasks, thereby enhancing resource allocation. Accessible via web browsers and utilizing cloud-based data processing, this innovative platform aims to provide a robust solution for advancing disaster response strategies.

S8.2: Modernizing Legacy IT Systems: Methods, Challenges, and Strategic Insights

Chaima Meradi, Sofiane Aouag, Assia BELOUCIF (University of Batna 2)

modernising legacy information technology (IT) systems is essential to improving efficiency, maintaining competitive advantage and driving business growth. This article provides an in-depth analysis of different approaches to modernising legacy systems, highlighting the importance of selecting appropriate methods and taking into account the common challenges faced by organisations during these transitions. The aim of this analysis is to compare different approaches to modernisation, identify best practices and offer concrete suggestions for overcoming difficulties.These objectives were achieved using a case study approach, ensuring an in-depth analysis of real-world modernisation projects in a wide range of areas. The research selected was chosen for its relevance, the quality of the data and its ability to provide valuable insights into the challenges of IT modernisation. Although this review was comprehensive, a number of difficulties and limitations were encountered, such as the diversity in scope and depth of existing research, which sometimes made it difficult to make accurate comparisons between each of the methodologies. In addition, specific organisational contexts and rapidly evolving technological landscapes have added additional levels of complexity, limiting the generalisability of the results. Nevertheless, this study provides essential insights and concrete recommendations to support organisations through the complex process of modernising IT systems, enabling them to make informed decisions tailored to their specific needs and situations.

Chairs: Dr. Youssef Elmir (ESTIN Bejaia, Algeria), Dr. Yousra Hlaoui (University of Tunis El Manar, Tunisia)

S9.1: A Decentralized GPS Positioning Consensus Based on Secure Image Watermarking for UAV Networks

Ferik Brahim, Laimeche Lakhdar, Abdalla Meraoumia (University of Larbi Tebessi Tebessa), Abdelkader Laouid (University of El Oued ) Ahcene Bounceur (University of Sharjah), Mostefa Kara (King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia) 

The popularity of Unmanned Aerial Vehicles (UAVs) in recent years has accelerated the demand for robust image security to secure communication and authenticate images or videos from usage by unauthorized parties who may seek access to sensitive information. Having unauthorized and deceptive communication, from interference in signals to data manipulation, will be a challenge.
This paper presents a novel approach to integrating GPS positioning with secure image capture and decentralized coordination for UAVs. Using El-Gamal encryption and Magic Number Fragmentation, GPS coordinates are securely encrypted, embed GPS data into captured images through a hybrid Discrete Wavelet Transform-Discrete Cosine Transform (DWT-DCT) watermarking technique. A consensus mechanism ensures decentralized and privacy-preserving computation of secure meeting points for UAVs. The system maintains high image quality, with superior PSNR and SSIM values, while demonstrating resilience against signal processing and geometric attacks. These findings affirm the method's ability to extract encrypted GPS coordinates reliably, enhancing UAV network security.

S9.2: Velocity-Based Detection of Unusual Movements for Enhanced Home Security

Wissam Benlala, YAZID Mohand ( University of Bejaia) 

Caring for individuals diagnosed with Alzheimer’s disease presents unique challenges, particularly in ensuring their safety at home. Memory impairments, a common symptom of Alzheimer’s disease, can lead to hazards such as unintentionally leaving principal doors and windows open, increasing vulnerability to intrusions. This paper proposes a novel intrusion detection system designed for smart homes, primarily utilizing motion sensors to protect patient privacy. The system analyzes movement velocities between strategically placed sensor pairs to monitor patient movements and identify unusual activities indicative of potential intrusions. We validate the system using the Technology Integrated Health Management (TIHM) dataset, providing insights into the movement patterns of Alzheimer’s patients within a smart home setting. By comparing movement velocities among patients, we identify variations in their patterns. Statistical analysis confirms the effectiveness of our approach in successfully identifying intrusions, thus enhancing the safety of Alzheimer’s patients at home.

S9.3 A Privacy-Preserving E-Voting System Using Blockchain and Linkable Ring Signatures

Ahmed-Sami Berkani, Hamouma Moumen, Saber Benharzallah, Youssouf Achouri (University of Batna 2), Mohand Tahar Kechadi (University College Dublin, Ireland)

In this work we propose a privacy-preserving, blockchain-enabled e-voting system that integrates Linkable Ring Signatures (LRS) to ensure a secure, scalable, and privacy-preserving solution for modern e-voting, providing a framework for trusted and transparent electoral processes in the digital age. Our system ensures voter anonymity while preventing double voting through the use of a linkability tag, while blockchain technology is employed to provide an immutable and transparent ledger for vote recording. The InterPlanetary File System (IPFS) is utilized for decentralized and GDPR-compliant storage of public keys. Performance analysis shows scalability challenges for LRS in large elections, but the system remains efficient for small scale use. Future work will focus on optimizing scalability for larger deployments.

S10.1: Human Activity Recognition in Enhancing Healthcare for Aging Populations: Challenges, Innovations, and Future Directions

Mohamed abderrahmen Boulahia, Saber BENHARZALLAH, Faiza Titouna, Tahar Dilekh (University of Batna 2) 

The aging global population poses significant challenges to healthcare and social systems. Addressing age-related health issues and fostering healthy aging requires innovative solutions. Researchers have increasingly focused on deep learning and Internet of Things (IoT) technologies to offer personalized, proactive, and preventive care for the elderly, aiming to enhance their quality of life and independence. This paper introduces essential concepts and methodologies used in the application of deep learning and IoT for advancing healthy aging. It explores the significance of accurately monitoring and understanding daily activities, with a particular focus on human activity recognition (HAR) as a critical component in assessing overall well-being among the aging population.
However, most current studies continue to emphasize achieving high accuracy in machine learning models, despite already reaching substantial levels of precision. This paper argues for the need to shift focus towards developing adaptable models capable of recognizing new and varied activities. Additionally, it highlights the importance of recommendation systems in healthcare, which should extend beyond mere activity recognition to provide actionable advice and preventive measures for the elderly.
This study also addresses the gap in the literature by providing a comprehensive review of existing approaches and identifying key research directions that should be prioritized in the field. By signaling the lack of comprehensive studies offering in-depth insights like this paper, it underscores the necessity of moving beyond current achievements towards more dynamic and context-aware solutions in HAR technologies.

S10.2: Comprenhensive Review of Smart Healthcare Recommendation Systems Using Deep Learning

Soumeya Bouslah, Saber BENHARZALLAH, Moumen Hamouma (university of Batna 2), Laid KAHLOUL (University of Biskra)

As smart cities advance and increasingly use data from Internet of Things (IoT) sources to enhance urban environments, smart healthcare has become a key focus area. This paper investigates the integration of deep learning techniques in Smart Healthcare Recommendation Systems (SHRSs), presenting a foundational taxonomy, that categorizes these systems into four distinct types: Content-Based Filtering (CBF), Collaborative Filtering (CF), Context-Aware Filtering (CAF), and Hybrid Filtering (HF). These approaches, implemented through Deep Learning, enable advanced data processing, pattern recognition, and personalization. Additionally, the paper introduces a new classification based on the nature of intervention, dividing SHRSs into preventive, diagnostic, treatment-focused, and hybrid categories. This dual classification fills a critical gap in the comparative analysis of deep learning-driven healthcare, which demonstrates the efficacy of these systems in providing personalized health advice, leveraging real-time sensor data and optimizing the handling of complex health datasets. Furthermore, this research outlines future work and potential directions for advancing SHRSs, offering insights into the application and impact of sophisticated recommendation techniques within the context of urban health management.

S10.3 A Robust Medical Image Watermarking Approach Based on K-Level Clustering and Asymmetric Encryption

Mohammed Elhabib Kahla, Mounir Beggas, Abdelkader Laouid (University of El Oued), Ahcene Bounceur (King Fahd University of Petroleum and Minerals, Saudi Arabia), Mostefa Kara (University of Sharjah, UAE)

The increasing use of digital medical images in healthcare has raised significant concerns regarding their security, privacy, and integrity during transmission and storage. Unauthorized access, tampering, and image manipulation pose severe risks, potentially leading to compromised diagnoses and patient safety issues. This paper presents a novel asymmetric image encryption and watermarking scheme aimed at securing medical images. The proposed scheme integrates cryptographic encryption and digital watermarking to protect sensitive patient information and ensure the authenticity and integrity of medical images. Specifically, the scheme encrypts patient data using asymmetric encryption and embeds the encrypted data into the medical image as a QR code watermark. This watermark is securely embedded into the least significant bits of the image pixels using a clustering algorithm, ensuring resilience against common image distortions, such as noise, rotation, and compression. Experimental results demonstrate the robustness of the proposed scheme, as evidenced by high PSNR and SSIM values, indicating minimal impact on image quality while ensuring strong security. The proposed method offers an efficient solution for enhancing the security of medical images in healthcare systems.

S11.1: Predicting Student Success in Mathematics Using Machine Learning Techniques: A Comparative Study of Algorithms.

MESTOUR Hareth, ZOUAOUI Samia (University of Batna 2), ZOUAOUI NourElhouda (University of Biskra) 

Achieving low results in mathematics is a real and urgent concern, as mathematics is the foundation of science and technology, and any weakness in it reflects negatively on the progress of society in various ways. Our study aims to evaluate the impact of different methods of studying mathematics on students' educational success and attainment. We employed three different machine learning models to analyze data related to mathematics study techniques and student outcomes. These methods included: support vector machine (SVM), A K-nearest neighbors (KNN), and Linear Regression (LR). The results showed that KNN provided the best accuracy rate with 97.6\% and an F1-score of 96.7\%. These results highlight the significant contribution of forming study groups and trying to cooperate with each other, this factor is considered important and influential in the success of students in mathematics.

S11.2: Analyzing Swarm Robotics Approaches in Natural Disaster Scenarios: A Comparative Study

Aicha HAFID, Riadh HOCINE, Lahcene GUEZOULI (University of Batna 2)

Natural disasters such as earthquakes, volcanic eruptions, tsunamis, and wildfires pose significant challenges for human intervention due to their unpredictable nature and the hazardous environments they create. In recent years, swarm robotics has emerged as a promising approach to address these challenges. Swarm robots, inspired by the collective behavior observed in nature, operate autonomously and collaboratively, offering unique advantages in disaster scenarios. These robots can rapidly deploy, adapt to dynamic environments, and provide extensive coverage, making them particularly suited for search and rescue missions, environmental monitoring, and infrastructure assessment. This paper presents a comprehensive review and comparative analysis of various approaches utilizing swarm robotics to tackle natural disasters. We explore key methodologies, including decentralized coordination algorithms, communication protocols, and adaptive navigation strategies, and evaluate their effectiveness in real-world disaster scenarios. Through a detailed examination of existing literature and case studies, we highlight the strengths and limitations of each approach, providing insights into their practical applications and potential for future developments. The findings underscore the critical role of swarm robotics in enhancing disaster response capabilities, offering scalable, resilient, and efficient solutions for mitigating the impact of natural disasters. We conclude with recommendations for future research directions aimed at overcoming current limitations and advancing the field toward fully autonomous and reliable swarm-based disaster management systems.

S11.3 Chemotaxis Bio-Inspired Clustering Algorithm to Develop Self-Organized UAV Topologies with Adaptive Behavior

Nedjma DJEZZAR (University of Batna 2)

This paper proposes a distributed and self-organized dynamic clustering scheme for Unmanned Aerial Vehicle (UAV) networks. The model is bio-inspired by the chemotaxis behavior of bacteria and uses light signals to dynamically adapt the movement of cluster follower nodes toward cluster heads. The proposed algorithm relies only on tumble and run movements, without requiring geographic data or complex computations. The results demonstrate the emergence of self-organized clustering behavior that adapts to the constant motion of nodes. Performance evaluations show that the optimal number of clusters achieved by the model outperforms existing bio-inspired approaches, such as Glowworm swarm intelligence schemes.

S11.4: Enhancing Proximal Policy Optimization in the Pendulum-v1 Environment through Clustering-Based State Space Simplification

Abid Mohamed Nadhir, Mounir Beggas, Abdelkader Laouid (University of El Oued), Mostefa Kara (King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia), Ahcene Bounceur (University of Sharjah, UAE)

This study aims to enhance the performance of the Proximal Policy Optimization (PPO) algorithm in continuous control tasks, particularly in the Pendulum-v1 environment. We introduce a clustering-based approach to simplify the state-action space, improving learning efficiency and maximizing cumulative rewards. The state-action pairs are clustered using k-means, emphasizing regions with higher rewards to guide the agent’s policy optimization. Experimental results demonstrate that clustering significantly improves PPO’s performance, with the 10-cluster configuration yielding the best outcomes in terms of reward consistency and efficiency. After tuning, the clustering-enhanced PPO achieves near-optimal performance, outperforming the baseline PPO by a wide margin.

S12.1: Real-Time Illuminance Assessment Web Application for Energy Efficiency Systems

Seniguer Abderraouf (Bordj Bou Arreridj University), Aouache Mustapha (CDTA, Algeria), Iratni Abdelhamid (Bordj Bou Arreridj University) 

The growing emphasis on energy efficiency in building management has led to increased adoption of daylight harvesting systems, which optimize the exploitation of sunlight to minimize the energy consumption of artificial lighting. Accurate lighting assessment is crucial for the effective implementation of these systems.
This paper presents a mobile-friendly web application designed to comprehensively assess lighting conditions in daylight harvesting systems. The application leverages smartphone sensors and deep learning technology to measure illuminance levels, providing real-time data and analytics to facility managers and building occupants. Key features include user-friendly interfaces for data visualization, integration with building management systems, and automated reporting functions. Preliminary field tests demonstrate the potential of the app to improve energy savings and optimize occupant comfort by facilitating the precise calibration of daylight-responsive lighting controls. The proposed solution represents a cost-effective and accessible tool for advancing sustainable building practices and supports the broader adoption of smart lighting technologies.

S12.2: Strategies for Service Placement in Fog Computing: Approaches and Challenges

Hanene Merouani, Sonia-Sabrina Bendib, Hamouma Moumen (University of Batna 2), Imene Achouri (University Badji Mokhtar, Annaba, Algeria)

Fog Computing has emerged as a crucial paradigm to address the limitations of traditional cloud computing, particularly in terms of latency, bandwidth, and real-time processing requirements. The placement of services within fog computing environments is a complex and vital process, directly impacting system performance, resource utilization, and Quality of Service (QoS) for various applications. This paper provides a comprehensive review of the state-of-the-art strategies for service placement in Fog Computing, categorizing them into four main types: mathematical programming, heuristics, metaheuristics, and machine learning. Each approach is analyzed in terms of its optimization objectives, strengths, and limitations. Additionally, we discuss the key challenges inherent in Fog Computing environments, such as resource variability, dynamic network conditions, and scalability. Finally, the paper explores potential avenues for future research, particularly the development of hybrid strategies that integrate the strengths of multiple approaches to address the evolving demands of Fog Computing environments.

S12.3 MCASM: A Framework Cloud_Mashup based on the concept of Assisted Search Model for Managing Data as a Service (DaaS).

Mourad BELABED, Abdeslem DENNAI, Nabil MOUSSAOUI, Younes KHAIR (TAHRI Mohamed University Bechar, Algeria)

New web technologies have made it easier to create, use, share and recycle online resources over time. Cloud computing provides the ability to create multiple applications and software systems to satisfy complex user needs while improving the utilization of shared resources and information. The concept of “mashups” was created through this technological advancement, which combines datasets and services from different sources into a single website. This makes cooperation between websites and web services easier. However, the abundance of data available poses challenges, including sifting through irrelevant information due to the vast volume of online data and the criteria used for searching Thus, the organization and management of results data are not taken into consideration. To tackle these issues, this article proposes the use of cloud Mashup applications employing an Assisted Search Model (ASM)[1]. By querying DaaS-Cloud (Data as a Service-Cloud) services, with the possibility of integrating other Clouds by applying the notion of InterCloud, this approach termed MCASM, enables the creation of cloud mashups that utilize ASM concepts. This service provides users with access to a broad spectrum of data sources and enhances the relevance of data selection based on specified parameters.

S13.1: Image Encryption via QSD Decomposition and Pairing Functions: A Novel Approach

Oussama Harkati, Lemnouar NOUI, Assia BELOUCIF (University of Batna 2) 

In today's digital world, information security is more important than ever. Cryptography, the science of encrypting data, plays a vital role in protecting sensitive information. However, with the development of attack threats and computational power the implementation of new cryptographic techniques and strategies becomes crucial to protect information. In this context, this paper presents a novel approach to encrypt grayscale images based on integer decomposition, called quasi-square decomposition, Cantor's pairing function and involutory matrix. Experimental results and security analysis demonstrate the effectiveness of the proposed approach, providing a robust solution for data protection.

S13.2: Ciphertext-Independent and Unbounded Depth RSA-Based Updatable Encryption Scheme

Mostefa Kara (King Fahd University of Petroleum and Minerals (KFUPM), Dhahran,Saudi Arabia), Abdelkader Laouid (El Oued University), Ahcene Bounceur (University of Sharjah, UAE), Mohammad Hammoudeh (Manchester Metropolitan University,UK), Abid Mohamed Nadhir (El Oued University), Kahla Mohamed El Habib (El Oued University)

We introduce a novel updatable encryption scheme based on the RSA cryptosystem to address the challenges of key rotation in secure communication systems. Updatable encryption allows ciphertexts encrypted under an old key to be transformed into ciphertexts under a new key without decrypting the plaintext, thereby enhancing security without compromising data integrity or confidentiality. Our RSA-based scheme leverages the mathematical properties of RSA to enable efficient and seamless key updates. We provide a security analysis demonstrating that our scheme is not vulnerable to information leaking. Experimental results indicate that our method offers a practical and efficient solution for large-scale systems that require regular key updates, with minimal impact on computational resources and performance.

S14.1: Application of Dataset Pruning and Dynamic Transfer Learning on Vision Transformers for MGMT Prediction on Brain MRI Images

Norelhouda Laribi, Djamel Gaceb, Fayçal Touazi, Abdellah Rezoug (University M'Hamed Bougara Boumerdès, Algeria) 

Deep learning models, in particular vision transformer, are becoming more dependent on large and representative datasets. Their application on 3D medical images also has additional challenges. In this context, two primary solutions are used : either creating sub-volumes and then slicing them for patch embedding, or generating 2D slices from the 3D data before patch embedding. It is important to efficiently convert 3D images into 2D slices, removing less useful ones, in order to optimize computational resources. When analyzing brain MRI images, glioblastoma lesions present a significant challenge in identifying critical regions for MGMT promoter methylation, as these regions are typically small and difficult to detect. To address this challenge and improve the cost-effectiveness of 3D Brain MRI imaging, This paper propose a novel approach based on Dataset Pruning, leveraging Transfer Learning with Vision Transformers (MobileViTV2) and tracks prediction uncertainty across multiple epochs to evaluate and prune less informative images . The results show that high accuracy and precision can be achieved by minimizing noise and removing less relevant images, enabling the model to focus on the most critical data.

S14.2: Transfer Learning Method Towards Lung And Colon Cancers Automated Analysis In Histopathological Images

Marwen SAKLI, Chaker ESSID1, Bassem BEN SALAH1 (University of Carthage, Tunisia) , Mohamed Habib AGREBI (Indiana University-Purdue University Indianapolis, Indianapolis, United States) and Hedi SAKLI (EITA Consulting, Montesson, France). 

Lung and colon cancers are two of the top causes of death and morbidity in humans. They may grow simultaneously in organs, having a deleterious influence on human life. If cancer is not detected early, it is likely to spread to both of those organs. In this research, transfer learning method based on a numerous model is presented and compared to atomate lung and colon cancers diagnosis in histopathological images. The LC25000 dataset is commonly used for this task. In fact, it includes 25000 histopathological images belonging to 5 distinct classes which are colon adenocarcinoma, benign colonic tissue, lung adenocarcinoma, lung squamous cell carcinoma, and benign lung tissue. For the transfer Learning method, the used models which were compared are EfficientNetsV2 from B0 to B3, and Small, MobileNetV2, ResNetV2, and Xception. The best results were reached when Employing EfficientNetV2B1. The scored accuracy, F1-score, AUC are equal to 1. Moreover, the loss is 6.06.10-5, respectively. Furthermore, the accuracy, F1score, AUC, Sensitivity, Specificity, attained 1 for the totality of the 5 classes. The proposed method with a transfer learning approach is extremely efficient for automated lung and colon cancer diagnosis, outperforming several current methods. This method might help doctors provide more accurate diagnoses and improve patient outcomes

S14.3: Application of Neural Network-Based U-Net Architecture for COVID-19 Region Detection in CT Scan Images

Mohamed Habib AGREBI, Faouzi BEN MASSAOUD (Indiana University-Purdue University Indianapolis, United States), Chaker ESSID, Marwen SAKLI (University of Carthage, Tunisia) 

Through this work, we have proposed a deep CNN architecture based Unet model, to automatically segment thoracic CT scans images. At first, the database has been augmented using preprocessing algorithms such as rotations and filtering. The second step, we have proposed a deep CNN architecture based two models: Unet and Segnet. High performance scores including accuracy, F1-score have been reached. Moreover, the Unet proved to be efficient to segment different regions such as normal lungs, glass representing Covid-19 and background regions. Results quantification have been reached to 97.98% for the train and 95.49% for the validation. The segmented tested images represent an error too close to zero, in comparison to the ground truth. The proposed segmentor system shows to be reliable, achieving an IOU value equal to 92% and 87%, an MSE equal to 0.08 and 0.13 for the validation and test processing respectively. Thus, our proposed system could be a good candidate to help doctors for fast covid19 diagnosis.

S15.1: Text Entailment Based Evaluation Approach for Generating Labelled Knowledge Graph from Textual Sources

Melyara MEZZI, Messaouda FAREH (University of Blida 1) 

The increase in textual data across various domains has prompted the need for structured formats to take advantage of the knowledge contained in these sources. Knowledge Graphs serve as an effective tool for organizing and extracting insights efficiently. However, constructing a Knowledge Graph from text presents formidable challenges due to the unstructured nature of textual data, which requires advanced Natural Language Processing techniques to parse and infer semantic relationships accurately. To evaluate the structure of the Knowledge Graph proposed in this paper, an approach based on Textual Entailment is introduced to assess the consistency and precision of Knowledge Graphs. This method leverages Textual Entailments to determine whether the textual data support the inferred relationships within the Knowledge Graph, thereby ensuring that the constructed graph faithfully represents the underlying knowledge.

S15.2: RE-SciBERT: An Entity-Enriched Language Model to enhance Biomedical Relation Extraction

Aicha Aid, Rima Azzoune, Imane Abdellaoui, Ahcene Haddouche (Bouira University)

Relation extraction is a key task in biomedical natural language processing (NLP), aiming to identify and extract relationships between entities across various sources, including clinical texts, research papers, and other biomedical literature. Recently, transformer-based models have set new benchmarks across various NLP tasks. However, relation extraction poses unique challenges, as it depends on understanding both the broader context and the specific interactions between targeted entities. In this paper, we introduce RE-SciBERT, an entity-enriched language model that leverages a modified SciBERT architecture, enhanced with entity embeddings, to capture nuanced relationships between biomedical entities. By integrating explicit entity information directly into the transformer model, we aim to improve the F1 scores of relation extraction tasks compared to standard transformer-based methods. We evaluate our model on benchmark datasets, demonstrating that the entity-enrichment strategy significantly boosts performance, particularly in scenarios involving complex or ambiguous entity relations. The results highlight the potential of transformer-based models with specialized entity embeddings in advancing the task of biomedical relation extraction.

S16.1: Security Considerations for Internet of Things (IoT): Exploring Quantum Computing Solutions

Youssouf Achouri*, Rima Djellab*, Khaled Hamouid**, Ahmed-Sami Berkani* (* University of Batna 2) (**ESIEE Paris, University of Gustave Eiffel, France) 

The rapid expansion of Internet of Things (IoT) devices has driven significant advancements across various sectors, including healthcare, smart cities, industrial automation, and home automation. However, this proliferation introduces numerous security vulnerabilities that compromise data privacy, integrity, and availability. Traditional security mechanisms often prove inadequate in IoT environments due to limited computational resources, heterogeneity, and scalability issues. This paper provides a comprehensive overview of IoT security considerations, highlighting essential security measures and the transformative potential of quantum computing. We examine fundamental security requirements for IoT systems and explore how quantum computing can enhance security through quantum key distribution (QKD) and post-quantum cryptography. Our analysis aims to contribute to the development of resilient and secure IoT systems capable of withstanding emerging threats.

S16.2: Comparative study of quantum transpilers: evaluating the performance of qiskit-braket-provider, qBraid-SDK, and pytket extensions

Mohamed Messaoud Louamri, Nacer eddine Belaloui, Abdellah Tounsi, Mohamed Taha Rouabah (University of Constantine 1, Algeria)

In this study, we conducted a comprehensive evaluation of several quantum transpilers, including the qiskit-braket-provider, the qBraid-SDK, and the pytket extensions, focusing on critical metrics such as correctness, failure rate, and transpilation time. Our results demonstrate that the qiskit-braket-provider exhibits superior performance achieving a remarkably low failure rate of 0.2 percent. the qiskit-braket-provider utilizes a combination of one-to-one transpilation and gate decomposition for unsupported gates, enhancing transpiler capabilities and speed. The qBraid-SDK offers a more generalized approach, suitable for transpilation across multiple SDKs, albeit with slower performance compared to the qiskit-braket-provider. The pytket extensions, while fast, exhibit limitations in handling complex circuits due to their one-to-one transpilation scheme. We also provide recommendations for future development, advocating for the adoption of the method employed by the qiskit-braket-provider to enhance transpiler capabilities and speed. This study contributes to the growing body of knowledge in quantum transpiler benchmarking, fostering interoperability and guiding the development of quantum computing applications in a diverse hardware and software landscape.

S16.3: Enhancing Quantum Search: The Potential of Partitioned Grover’s Algorithm

Ahcene Bounceur, Saadat M. Alhashmi, Mohamed Abdalla Nour and Huseyin Seker (University of Sharjah, UAE)

Grover's Algorithm is a prominent quantum search algorithm in the rapidly advancing quantum computing study. By exploiting superposition and quantum parallelism, it has the potential to outperform classical brute-force approaches. Nevertheless, the current practical application of quantum technology in decrypting intricate passwords is restricted by its early stage, namely the restricted qubit quantities and short coherence durations. This work presents an innovative distributed quantum computing paradigm that addresses these difficulties by partitioning the password search space among several quantum processors. Each processor executes Grover's Algorithm on a subset of the search space, minimizing the computational complexity and quantum resource requirements for each individual processor. The partial findings are subsequently transmitted to a central classical computer, which amalgamates them to obtain the ultimate password. This distributed methodology not only allows for practicality with current quantum systems but also provides a scalable paradigm that can adapt to the progress of quantum computer technology. By decomposing the quantum brute-force search problem into smaller, more controllable elements, this approach can significantly decrease the time complexity of password retrieval, which has significant consequences for cryptography and cybersecurity.