CH1902033 - Nguyen Trong Minh Hong Phuoc
In today's context, with the explosion of Information Technology (IT) and its auxiliaries, our lives are undergoing a significant transformation. This relentless development in IT sectors poses a major challenge: how to maintain competitiveness and adaptability. In this context, data digitization is becoming a top priority. Simultaneously, it opens up significant opportunities for sophisticated attackers. Attackers and information thieves are becoming increasingly cunning, always seeking unauthorized access to data storage systems. It's noteworthy that traditional rule-based detection systems are gradually becoming weak and even outdated when faced with the sophistication of modern attacks.
With the idea of combining the advantages of two methodological groups in the IDS problem: Machine Learning-based and Deep Learning-based, this topic selects the combination to create a feature template of network activity data with high discrimination as follows: A feature template with high discrimination, selected and contributing to distinguishing network attacks and normal states, less affected by noise. Learning from features based on DL techniques allows learning important features from network data. Experiments conducted using the dataset of the University of New Brunswick, Canada (NSL-KDD) show that the proposed method outperforms existing published methods.
Best regards.
For more information, please visit: https://fit.uit.edu.vn/index.php/tin-tuc/goc-hoc-tap/6611-phat-hi-n-xam-nh-p-s-d-ng-xgboost-d-l-a-ch-n-tinh-nang-v-i-logarithmic-autoencoder-va-bidirectional-lstm
Đông Xanh - Media Collaborator at University of Information Technology
