Abstract: The data is an asset for oil companies. Both data encryption and data desensitization are aimed at data security strategies. However, current data desensitization technology can only be implemented for text. There is a lack of image data desensitization technology that is widely available in the oil industry. Aiming at the shortcomings of the current desensitization technology and the characteristics of curve image data in oil and gas field exploration, this paper draws on the deep learning image restoration method and proposes a set of desensitization technology based on context encoder. It made up for the defect that the industry could not desensitize image data. After detailed testing of the method, a good desensitization effect was achieved. Compared with the traditional method, the result proves that it can achieve more obvious effect and more reliable desensitization effect in the special data of curve image.