Open Access

ARTICLE

AI-Driven Identification of Attack Precursors: A Machine Learning Approach to Predictive Cybersecurity

Abdulwahid Al Abdulwahid^*

Department of Computer and Information Technology, Jubail Industrial College, Royal Commission for Jubail and Yanbu, Jubail Industrial City, 31961, Saudi Arabia

* Corresponding Authors: Abdulwahid Al Abdulwahid. Email: or

(This article belongs to the Special Issue: Artificial Intelligence Methods and Techniques to Cybersecurity)

Computers, Materials & Continua 2025, 85(1), 1751-1777. https://doi.org/10.32604/cmc.2025.066892

Received 19 April 2025; Accepted 14 July 2025; Issue published 29 August 2025

Abstract

The increasing sophistication of cyberattacks, coupled with the limitations of rule-based detection systems, underscores the urgent need for proactive and intelligent cybersecurity solutions. Traditional intrusion detection systems often struggle with detecting early-stage threats, particularly in dynamic environments such as IoT, SDNs, and cloud infrastructures. These systems are hindered by high false positive rates, poor adaptability to evolving threats, and reliance on large labeled datasets. To address these challenges, this paper introduces CyberGuard-X, an AI-driven framework designed to identify attack precursors—subtle indicators of malicious intent—before full-scale intrusions occur. CyberGuard-X integrates anomaly detection, time-series analysis, and multi-stage classification within a scalable architecture. The model leverages deep learning techniques such as autoencoders, LSTM networks, and Transformer layers, supported by semi-supervised learning to enhance detection of zero-day and rare threats. Extensive experiments on benchmark datasets (CICIDS2017, CSE-CIC-IDS2018, and UNSW-NB15) demonstrate strong results, including 96.1% accuracy, 94.7% precision, and 95.3% recall, while achieving a zero-day detection rate of 84.5%. With an inference time of 12.8 ms and 34.5% latency reduction, the model supports real-time deployment in resource-constrained environments. CyberGuard-X not only surpasses baseline models like LSTM and Random Forest but also enhances proactive threat mitigation across diverse network settings.

---

Keywords

---