Prevention of cyber attacks and real-time social media spam detection and sentiment analysis using recurrent self-adaptive windowing approach

Social media platforms, with billions of users, have become an integral part of our lives. The projected exploration tackles the long-standing issue of spam detection on social media platforms by developing a novel approach that combines pre-processing, recurrent neural networks (RNNs), and optimisation techniques. The exploration makes four key contributions, namely, the development of an efficient pre-processing pipeline for preparing e-mail data, the application of an RNN with a soft attention mechanism for accurate spam classification, the integration of self-adaptive windowing to address concept drift, and the optimisation of the RNN using weighted chimp optimisation. The latter two innovations enable the model to adapt to changes in spamming tactics over time and improve its overall performance. As a result, the research achieves a high accuracy rate of 99.3% and outperforms traditional methods in terms of precision, recall, and F1-score. While these results are impressive, the limitations of relying solely on accuracy are acknowledged, and the model's performance on imbalanced datasets will be further explored in subsequent sections to provide a more comprehensive understanding of its effectiveness.