Current Status and Challenges and Future Trends of Deep Learning-Based Intrusion Detection Models

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Machine Learning

Background:

• Deep learning (DL) is increasingly vital for cybersecurity intrusion detection.
• Research in DL-based intrusion detection models is rapidly advancing.
• Effective intrusion detection relies on robust datasets and preprocessing.

Purpose of the Study:

• To provide a comprehensive overview of datasets used in DL-based intrusion detection research.
• To summarize prevalent data preprocessing and feature engineering techniques.
• To review various DL models and their applications in cybersecurity.

Main Methods:

• Literature review of datasets, preprocessing methods, and feature engineering.
• Analysis of seven DL models: deep autoencoders, deep belief networks, deep neural networks, convolutional neural networks, recurrent neural networks, generative adversarial networks, and transformers.
• Inclusion of large-scale predictive models like BERT and GPT series for intrusion detection.

Main Results:

• Identified widely utilized datasets and common preprocessing/feature engineering techniques.
• Examined architectures and applications of seven distinct DL models.
• Demonstrated the efficacy of transformer-based models (BERT, GPT) in intrusion detection.

Conclusions:

• DL-based intrusion detection systems show significant promise for enhanced accuracy and efficiency.
• Future research should focus on identified key areas to address evolving cybersecurity threats.
• Advanced models like transformers are crucial for adapting to dynamic threat landscapes.