Realtime keylogger detection and alert

Keyloggers are among the most insidious forms of malware, capable of covertly recording every keystroke a user makes and transmitting sensitive information such as login credentials, banking details, and private communications to malicious actors. Traditional signature-based detection methods, while effective against known threats, fail to recognize polymorphic or stealthy keyloggers that continuously evolve to bypass security measures. To address this gap, this project proposes a hybrid approach that combines machine learning, deep learning, and blockchain technologies for real-time keylogger detection and alerting. The methodology begins with synthetic dataset generation that simulates real-world system behavior through parameters such as CPU utilization, memory consumption, keystroke frequency, and network activity. Logistic Regression and Random Forest classifiers provide strong interpretability and baseline accuracy, while a Long Short-Term Memory (LSTM) network captures sequential dependencies in time-series data, enabling robust anomaly detection. The system achieves high detection accuracy with minimal false positives, validating the suitability of deep learning for behavioral profiling. To enhance transparency and trust, detection events and model updates are immutably logged onto a permissioned blockchain. This ensures tamper-proof auditability, decentralized model improvement through federated learning, and protection against single points of failure. Overall, the proposed system offers a scalable, secure, and adaptive framework for detecting both known and unknown keyloggers. Beyond strengthening cybersecurity, the approach establishes a foundation for extending AI–blockchain integration to broader malware detection challenges across critical sectors.