Safety is vital and surveillance technology required to achieve this is becoming increasingly challenging and complex. Smart enterprises, today, do not require obsolete, analog surveillance systems but new, dynamic automated approaches that provide strong safety and high business value from day one. Surveillance systems are an essential part of securing your home or business. However, the traditional methods of surveillance need a continual human observation though dozens of monitors in real-time. This may result in missing vulnerable situations due to fatigue, lack of concentration, or loss of key information in the surveillance videos. Therefore, an intelligent surveillance system is critically needed to automatically detect security threats in a stream of videos without manual intervention. This project presents NoxEye, a lightweight AI-powered threat detector for intelligent surveillance cameras, which can be deployed on-site at the edge. The goal of NoxEye is to minimize communication delays, which is essential to perform sensitive and mission-critical tasks such as thread detection using surveillance cameras. NoxEye is organized into two parts executing on centralized servers on the cloud, as well as locally on the surveillance cameras. we developed a user-friendly interface on top of both CNN and Faster Region-CNN to allow users to interact with the threat detector system conveniently at the camera and cloud sides. Third, a novel motion detection module is proposed for detecting moving objects in surveillance videos in real-time. The developed module is integrated seamlessly with both the camera and cloud sides.