This paper presents a real-time framework for detecting, recognizing, and tracking multiple faces in video streams, made for security and surveillance use cases. The proposed method integrates Kernelized Correlation Filter (KCF) for appearance-based tracking with Histogram of Oriented Gradients (HOG) for efficient feature extraction and accurate localization. After initial face detection, the method ensures continuous tracking through a hybrid approach that supports face re-identification and similarity score matching to maintain identity consistency across frames. This framework addresses key challenges such as occlusion, abrupt motion, and illumination variation by dynamically adapting tracking strategies based on accuracy and reliability metrics. Evaluation was conducted using benchmark datasets of WIDER FACE and the YouTube Faces Database. The results demonstrate strong performance, with high recognition precision (0.90), recall (0.89), and tracking accuracy measured by an MOTA score of 0.89. Overall, the system performs effectively in complex and unconstrained video environments, making it well-suited for real-world applications such as surveillance, identity verification, and behavioral analysis.