Scholarlink

<p>With the rapid advancement of technology, servers have become increasingly vulnerable to cyber threats, posing significant risks to valuable assets such as cloud infrastructure, IoT devices, and mobile applications. As cyberattacks escalate across various industries, the role of intrusion detection systems (IDS) in maintaining cybersecurity has become more critical than ever. Traditional (IDS) face substantial challenges in analyzing large volumes of operational data, often relying on recorded attack instances and anomaly detection, which may not suffice in the face of evolving threats. To overcome these limitations, recent advancements have focused on leveraging machine learning and deep learning-based (IDS). In this study, we introduce a multi-objective optimization-based hybrid method that integrates lightweight deep learning models to significantly enhance cybercrime detection accuracy. Our approach utilizes QR code images embedded with diverse datasets to train the models, capitalizing on the MobileNetV2 Convolutional Neural Network (CNN) for effective image processing and feature extraction. Post-training, feature extraction is applied to both the training images and the deep models. To refine the classification process, a genetic algorithm (GA) is employed to identify top-ranking features, utilizing accuracy, precision, recall, and F1-score to achieve a success rate of around 96%. With an F1-score of 0.86, sensitivity of 0.85, specificity of 0.90, precision of 0.88, and accuracy of 96.2%, the SVM classifier performed well. An accuracy of roughly 85.78% 60 was also attained by the KNN classifier, which was also tested. The overall approach highlights how well the lightweight deep learning model, MobileNetV2, extracts features from QR code images, which helps explain the remarkable accuracy rates that have been noted. The potential of lightweight deep learning models in conjunction with optimization strategies to improve cybersecurity protocols and lessen changing cyberthreats is demonstrated by this study.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing cybercrime detection: A hybrid deep learning approach for enhanced intrusion detection systems

  • 期刊论文

  • 发表日期：

    2025年4月10日

查看全文
  • Rima Alshaya,
  • Salim E. L. khediri

摘要

With the rapid advancement of technology, servers have become increasingly vulnerable to cyber threats, posing significant risks to valuable assets such as cloud infrastructure, IoT devices, and mobile applications. As cyberattacks escalate across various industries, the role of intrusion detection systems (IDS) in maintaining cybersecurity has become more critical than ever. Traditional (IDS) face substantial challenges in analyzing large volumes of operational data, often relying on recorded attack instances and anomaly detection, which may not suffice in the face of evolving threats. To overcome these limitations, recent advancements have focused on leveraging machine learning and deep learning-based (IDS). In this study, we introduce a multi-objective optimization-based hybrid method that integrates lightweight deep learning models to significantly enhance cybercrime detection accuracy. Our approach utilizes QR code images embedded with diverse datasets to train the models, capitalizing on the MobileNetV2 Convolutional Neural Network (CNN) for effective image processing and feature extraction. Post-training, feature extraction is applied to both the training images and the deep models. To refine the classification process, a genetic algorithm (GA) is employed to identify top-ranking features, utilizing accuracy, precision, recall, and F1-score to achieve a success rate of around 96%. With an F1-score of 0.86, sensitivity of 0.85, specificity of 0.90, precision of 0.88, and accuracy of 96.2%, the SVM classifier performed well. An accuracy of roughly 85.78% 60 was also attained by the KNN classifier, which was also tested. The overall approach highlights how well the lightweight deep learning model, MobileNetV2, extracts features from QR code images, which helps explain the remarkable accuracy rates that have been noted. The potential of lightweight deep learning models in conjunction with optimization strategies to improve cybersecurity protocols and lessen changing cyberthreats is demonstrated by this study.