Abstarct: Today, with the spread of cyber threats and various problems caused by intrusion into computer networks, many studies have been conducted on image encryption. Each of these researches has different strengths and weaknesses. The problem of most of these researches is the large amount of calculations and their average security. Therefore, time and security are two key factors for an encryption process that have been examined in this thesis. For this purpose, in the first step of the proposed method, we intend to use an efficient algorithm called YOLO algorithm to detect objects and also to detect the best object for stealth, using an algorithm called GLCM. In the next step, encryption is performed in the proposed object using a method baxsed on LSBMR. The proposed method in this thesis has two major advantages over previous methods. First, it has a strong security performance, along with a suitable embedding speed, which has been less considered in previous encryption systems. Then, special attention has been paid to maintaining maximum image quality. This process has been achieved thanks to the use of deep neural networks in object detection and selection of the appropriate object and the use of the LSBMR-baxsed method, which causes less changes in the image. Host images, like a face image, contain sensitive information and the slightest change will be noticeable to the viewer. The proposed method in this thesis also focuses on the protection of important areas and will prevent drastic changes in the image. ..The degree of improvement in the proposed method compared to other researches in this field has been reduced by more than 0.05% in the standard value of mean square error and increased by 2% in the standard value of PSNR.