AI-enhanced model for detecting cybersecurity threats on social media platforms

AI-enhanced model for detecting cybersecurity threats on social media platforms

The proliferation of cyber threats on social media requires the development of robust detection systems. X (formerly Twitter) is a prime example due to its extensive user base and real-time nature. Despite advancements in key enabling technologies such as Artificial Intelligence (AI), Machine Learning (ML), big data, blockchain, cloud computing, and the Internet of Things (IoT), the frequency and sophistication of cyberattacks escalate. Thus, cyber security remains a critical priority for X, as existing detection systems struggle to identify increasingly complex threats. This study addresses cybersecurity threats, the motivations behind them, and the primary challenges in detecting them on X. Current studies lack comprehensive evaluations of critical factors such as prediction scope, types of cyber security threats, feature extraction techniques, algorithm complexity, information summarisation levels, scalability over time, and performance measures. The surge in social media activities, particularly tweets, exacerbates the problem, making it imperative to develop more effective detection techniques. Traditional methods, including anomaly detection and rule-based approaches, are time-consuming, resource-intensive, and prone to inaccuracies. By contrast, AI, especially ML and DL, enhances the precision of cyber threat assessments. The research introduces a Subspace Random Ensemble Machine Learning Model (SREMLM) and a Voting Ensemble Deep Learning Model (VEDLM) designed for cyber threat detection and addressing cybersecurity challenges on social media. The findings demonstrate that the proposed SREMLM, and the proposed VEDLM, surpass individual DL and ML models in identifying cyber threats. Comparative analysis further confirms that this voting ensemble model, which incorporates multiple deep learning techniques, consistently delivers superior performance. The overall performance of the Voting Ensemble Deep Learning Model (VEDLM) exceeds the Subspace Random Ensemble Machine Learning Model (SREMLM), showcasing the advantage of using deep learning approaches. Furthermore, this research proposes a conceptual framework grounded in real-world datasets to enhance the practical applicability of the findings. The development of the novel proposed Ensemble ML, DL model advances threat detection capabilities, offering a pathway to bolstered security on the X platform. The study underscores the potential of ML, DL methods to revolutionise cyber threat detection, ensuring more robust and effective defense mechanisms against evolving cyber threats.

File reproduced in accordance with the copyright policy of the publisher/author.