Protecting the security of a maritime area is an important security interest of any state. In our project, we suggest deep learning based method for detecting and classifying ships based on their transmitted radar signals. We represent the signals by their spectrograms (time-frequency representation) and use them as images. The advantage of using images is that neural networks have been shown to be very efficient in image classification, and we have indeed achieved good results in the classification stage. Another problem we faced during the project was detecting and learning "new" ships that we had not been seen during the training stage, because we cannot assume that we know all the ships that will arrive to our area in the future. This problem, known as "open world recognition", has recently started to be investigated, and we adopted several solutions proposed in the literature to solve our problem. We managed to only detect a new ship but not learn it for future detection.