Recently, the challenges of security systems in protecting citizens from crime and disasters have become very difficult. While some organizations have developed fire detection security systems, incidents of intrusions, thefts, and hostage-taking in government and private agencies indicate the inadequacy of current security systems. In this paper, the security system combines the detection of people, weapons, and fires through the analysis of surveillance cameras and videos is presented. The system is built by using deep learning techniques. The system deals with the video frames and processes them using a set of digital filters. The system analyzes the tires and goes through the processes of detecting faces, detecting weapons, and detecting fires respectively. After that, the classification stage begins to identify people’s faces if they are outlaws and to identify the type of weapon. The deep learning models were applied in the classification process: Convolutional Neural Network (CNN), VGG-19 network, and GoogleNet with Inception module. The Inception network archives the highest accuracy for face recognition. The VGG network archived 95.5%, and 99.7% for weapon and fire detection respectively.