NLP-Driven Virtual Educator for Smart Teaching

Abstract

With the exponential growth of digitalization and data volumes, the cybersecurity threat landscape has become increasingly complex, amplifying the need for robust intrusion detection systems (IDS). Traditional IDS approaches often struggle with static architectures, requiring costly and frequent retraining to keep up with evolving threats. This study introduces an incremental, majority-voting IDS system that leverages machine learning to adapt to continuous network traffic streams without the need for extensive retraining. By integrating multiple machine learning algorithms—K-Nearest Neighbors (KNN), Logistic Regression, Bernoulli Naive Bayes, and Decision Tree classifiers—the system employs a collective decision-making approach to enhance detection accuracy and minimize false alarms in real-time.The proposed IDS framework is designed to handle large-scale, imbalanced network data, which is common in real-world environments. It offers enhanced adaptability by dynamically learning from new patterns, ensuring improved detection of both known and emerging threats. The ensemble method also reduces the risk of overfitting, making the system more reliable.Results from extensive simulations demonstrate that this multi-algorithm IDS outperforms traditional models in terms of accuracy, precision, and recall. Furthermore, the system's resilience to adversarial attacks and reduced retraining overhead make it a viable solution for modern, large-scale cybersecurity applications.