Traditional CCTV monitoring systems employ a person to manually monitor the screen to check for any intruders/trespassers. A step advanced to this would be employing deep models to check for any movement on the display. However, this is still very inefficient. This project develops a Deep CNN model to detect any trespassers. The model is executed in a manner which is different from the existing works. We create 3 classes, one for no people in the image, one for people who are known and the 3rd class for unknown people, who would be treated as intruders. In this manner, we can avoid a large class for which the model sends an alert, which would turn out to be a member of the house. A closed loop is introduced in a manner where we continuously update the model based on the user input of whether they can be included into the manner. If done so, then the model is additionally trained, and the data is included. This is further pushed into a web application so that users can control this operation remotely.