Preventing Cybercrime through Artificial Intelligence and Machine Learning in Education, Kenya

Abstract

The increasing reliance of educational institutions on digital systems has made them vulnerable to cybercrime, resulting in data breaches, financial loss, and disruption of learning activities. This study investigates the application of Artificial Intelligence (AI) and Machine Learning (ML) to prevent cyber threats in Kenyan educational institutions. A review of relevant literature guided the identification of suitable AI and ML algorithms, which were then tested using secondary datasets, including NSL-KDD (40 MB), PhishTank (15,000 URLs), and CICIDS2017 (24 GB), alongside simulated real-time cyber-attack logs. The datasets were split into 80% for training and 20% for testing, with cross-validation applied to prevent overfitting. Supervised learning models (Random Forest, Support Vector Machines) were used to classify known threats, unsupervised learning (K-means clustering) detected anomalous behaviors, and reinforcement learning optimized responses to dynamic threats. System performance was evaluated using accuracy, precision, recall, and false positive rate. Results showed that the reinforcement learning model achieved the highest effectiveness (95% accuracy, 96% recall, 4% false positive rate), while Random Forest also demonstrated high reliability in threat detection. The study highlights the ethical considerations of AI deployment, including privacy, bias, and responsible use, and recommends integrating hybrid AI models with human oversight to strengthen cybersecurity in educational institutions. These findings indicate that AI and ML provide robust, adaptive, and proactive solutions for preventing cybercrime in the education sector.