AI-BAHSI: A Hybrid Method of Artificial Intelligence-Behavioral Analysis and Hybrid Security Intelligence for Real-time Threat Detection and Mitigation on Wireless Access Points

Abstract

Wireless access point (AP) security faces significant challenges with the emergence of sophisticated attacks such as SSID Confusion (CVE-2023-52424), KRACK attacks, and advanced persistent threats. This research develops a hybrid AI-BAHSI (Artificial Intelligence-Behavioral Analysis and Hybrid Security Intelligence) method that integrates deep learning, ensemble machine learning, and federated learning for real-time threat detection and mitigation on wireless access points. The proposed method combines Convolutional Neural Network-Long Short Term Memory (CNN-LSTM) for pattern recognition, Random Forest-Support Vector Machine ensemble for threat classification, and federated learning for privacy-preserving security intelligence. Evaluation was conducted on a synthetic dataset that includes 15,000 normal traffic samples and 8,500 attack samples of various types. The results show that AI-BAHSI achieves a detection accuracy of 98.7%, a precision of 97.3%, a recall of 98.1%, and an F1-score of 97.7% with a false positive rate of only 1.2%. This method successfully detected zero-day attacks with a 94.6% confidence level and was able to automatically mitigate them in an average of 0.8 seconds. The main contribution of this research is the development of an adaptive security framework that can learn from new attack patterns in real time while preserving privacy through a federated learning architecture.