Zero-day exploits detection with adaptive WavePCA-Autoencoder (AWPA) adaptive hybrid exploit detection network...

Area of Science:

• Computer Science
• Cybersecurity
• Artificial Intelligence

Background:

• Existing anomaly detection systems struggle with accuracy, computational time, and adaptability for zero-day exploit detection.
• Zero-day exploits pose a significant threat due to their novelty and the limitations of current detection methods.

Purpose of the Study:

• Introduce a new probabilistic composite model to enhance zero-day exploit detection capabilities.
• Address the limitations of current anomaly detection systems in terms of accuracy, computational time, and adaptability.

Main Methods:

• Developed an Adaptive WavePCA-Autoencoder (AWPA) for denoising and dimensionality reduction.
• Introduced a Meta-Attention Transformer Autoencoder (MATA) for advanced feature extraction.
• Implemented Genetic Mongoose-Chameleon Optimization (GMCO) for efficient feature selection.
• Designed an Adaptive Hybrid Exploit Detection Network (AHEDNet) for dynamic ensemble adaptation.

Main Results:

• The proposed model achieved high accuracy (e.g., 0.9919 on dataset 2) and precision (e.g., 0.9968 on dataset 2).
• Demonstrated superior performance over existing models with significantly lower Hamming Loss on both datasets.
• Showcased improved detection of zero-day exploits with high accuracy and low false positives.

Conclusions:

• The novel probabilistic composite model significantly outperforms existing methods in detecting zero-day exploits.
• The integrated framework effectively addresses challenges in accuracy, computational efficiency, and adaptability.
• The proposed approach offers a robust solution for enhancing cybersecurity against emerging threats.