A hybrid model for detecting intrusion in cloud computing networks using artificial intelligence techniques

Authors

  • فادي غصنه أستاذ في قسم هندسة تكنولوجيا الاتصالات- كلية هندسة تكنولوجيا المعلومات والاتصالات – جامعة طرطوس- سوريا
  • ناجي محمد قسم هندسة تكنولوجيا الاتصالات – كلية هندسة تكنولوجيا المعلومات والاتصالات – جامعة طرطوس- سوريا
  • رنيم يونس قسم هندسة تكنولوجيا الاتصالات- كلية هندسة تكنولوجيا المعلومات والاتصالات-جامعة طرطوس-سوريا

Keywords:

Cloud Computing – Security – Deep Learning – DNN – Auto Encoder.

Abstract

With the rapid development of cloud computing technology, which has enabled the connection of numerous smart devices, facilitated easier data storage, exchange between systems, and allowed for more affordable access to services. The number of attacks and threats to cloud computing networks has simultaneously increased. Traditional solutions are no longer effective in protecting data and detecting or preventing threats in real time, including intrusion detection systems (IDS). This is where the importance of integrating artificial intelligence (AI) technologies into these systems lies. Advanced methods have been developed to enhance the performance of threat detection systems, such as machine learning and deep learning.

In this research, we conducted a statistical study on the CIC-IDS database, revealing several important characteristics of network traffic behavior. These characteristics justify the use of Auto Encoder (AE) technology as a crucial step in the attack classification process.  Following statistical analysis, the Deep Neural Network (DNN) algorithm was tested for attack detection. The results showed that it was only able to distinguish normal data traffic but could not detect the types of attacks present in the database. Therefore, a hybrid system was proposed, consisting of Auto Encoder technology as a first stage. This technology extracts important features, reduces data spacing, and detects anomalies by training it solely on normal data. The output of this technology serves as the input for the DNN algorithm, which is then used to classify the data traffic. The results demonstrated that the hybrid model performed well in classifying attack types, leading to improved intrusion detection systems.

Downloads

Published

2026-10-29