The progression of complexity in cyber-attacks requires the migration from reactive defenses to proactive defenses. This paper provides a design and assessment of an autonomous predictive threat detection system in cyberspace. The system employs machine learning, combined with real-time analysis, to predict and counter threats before they affect the system. The framework is based on a new ensemble learning algorithm that combines a deep neural network and a random forest classifier, striking a balance between high accuracy and a low false-positive rate. The framework was evaluated using the CICIDS2017 dataset, a real and extensive network traffic dataset that features a wide range of modern cyberattacks. The system was developed using Python as the programming language, along with Scikit-learn and TensorFlow, two prominent machine learning libraries. The result is that the autonomous system can detect different types of cyberattacks with an accuracy rate of over 98%, compared to traditional signature-based or individual machine learning-based detection methods. Autonomy of the system reduces human interaction, thereby enabling real-time and scalable cyber defense. The conclusion reached in this study provides a solid foundation for developing the next generation of predictive security solutions, particularly in the context of cybersecurity.
Anala Venkata Sai Abhishek (Tue,) studied this question.