Academic & Scientific Articles

Permanent URI for this communityhttp://dl.cerist.dz/handle/CERIST/3

Browse

Has files: YesSubject: search.filters.subject.Deep Learning

Search Results

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Multi-CNN Model for Multi-Classification of Cultural Heritage Monuments
    (CERIST, 2024-04) Djelliout, Toufik; Aliane, Hassina
    The use of convolutional neural networks (CNN) in the preservation of cultural heritage monuments, especially in conflict-affected regions such as Gaza, Ukraine, Iraq and others, represents a significant advancement in heritage conservation efforts. This paper presents an approach that uses a Multi-CNN model to classify images of cultural heritage monuments into various categories, encompassing period, monument type and location. By leveraging the capabilities of CNNs, this model demonstrates a high level of accuracy in categorizing heritage monuments based on multiple attributes. The study highlights the superior performance of the Multi-CNN model compared to other popular models such as DenseNet169, GoogleNet and MnasNet, highlighting its effectiveness in accurately classifying images of cultural heritage monuments in various dimensions. According to the evaluation results, the top-performing multi-CNN model achieves a classification accuracy of 94.52%, outperforming the single CNN models. The DenseNet196 model achieves 93.70% accuracy, the MnasNet model achieves 92.80% accuracy, and the GoogleNet model achieves 88.18% accuracy.
  • Thumbnail Image
    Item
    PReDIHERO – Privacy-Preserving Remote Deep Learning Inference based on Homomorphic Encryption and Reversible Obfuscation for Enhanced Client-side Overhead in Pervasive Health Monitoring
    (IEEE, 2021) Boulemtafes, Amine; Derhab, Abdelouahid; Ait Ali Braham, Nassim; Challal, Yacine
    Homomorphic Encryption is one of the most promising techniques to deal with privacy concerns, which is raised by remote deep learning paradigm, and maintain high classification accuracy. However, homomorphic encryption-based solutions are characterized by high overhead in terms of both computation and communication, which limits their adoption in pervasive health monitoring applications with constrained client-side devices. In this paper, we propose PReDIHERO, an improved privacy-preserving solution for remote deep learning inferences based on homomorphic encryption. The proposed solution applies a reversible obfuscation technique that successfully protects sensitive information, and enhances the client-side overhead compared to the conventional homomorphic encryption approach. The solution tackles three main heavyweight client-side tasks, namely, encryption and transmission of private data, refreshing encrypted data, and outsourcing computation of activation functions. The efficiency of the client-side is evaluated on a healthcare dataset and compared to a conventional homomorphic encryption approach. The evaluation results show that PReDIHERO requires increasingly less time and storage in comparison to conventional solutions when inferences are requested. At two hundreds inferences, the improvement ratio could reach more than 30 times in terms of computation overhead, and more than 8 times in terms of communication overhead. The same behavior is observed in sequential data and batch inferences, as we record an improvement ratio of more than 100 times in terms of computation overhead, and more than 20 times in terms of communication overhead.
  • Thumbnail Image
    Item
    Fault-tolerant AI-driven Intrusion Detection System for the Internet of Things
    (Elsevier, 2021-09) Medjek, Faiza; Tandjaoui, Djamel; Djedjig, Nabil; Romdhani, Imed
    Internet of Things (IoT) has emerged as a key component of all advanced critical infrastructures. However, with the challenging nature of IoT, new security breaches have been introduced, especially against the Routing Protocol for Low-power and Lossy Networks (RPL). Artificial-Intelligence-based technologies can be used to provide insights to deal with IoT’s security issues. In this paper, we describe the initial stages of developing, a new Intrusion Detection System using Machine Learning (ML) to detect routing attacks against RPL. We first simulate the routing attacks and capture the traffic for different topologies. We then process the traffic and generate large 2-class and multi-class datasets. We select a set of significant features for each attack, and we use this set to train different classifiers to make the IDS. The experiments with 5-fold cross-validation demonstrated that decision tree (DT), random forests (RF), and K-Nearest Neighbours (KNN) achieved good results of more than 99% value for accuracy, precision, recall, and F1-score metrics, and RF has achieved the lowest fitting time. On the other hand, Deep Learning (DL) model, MLP, Naïve Bayes (NB), and Logistic Regression (LR) have shown significantly lower performance.