Many services use passwords for personal authentication.In addition to passwords, there are also image authentication methods, but there is a risk of eavesdropping.There are some personal authentication systems that use hard-to-remember images to prevent this, but it's difficult to prevent screen recording attacks.

　This time, the research team has developed a personal authentication system "EYEDi" that applies an image processing filter to an image prepared by a user to generate a distorted image that only the person who knows the original image can understand, and presents it on the authentication screen.

　EYEDi applies an image processing filter to an image prepared by the user to generate a distorted image.Based on the memory, the user selects an image prepared by himself / herself from many distorted images and performs personal authentication.Since this distorted image cannot be restored and the intensity of the distortion can be adjusted, only a legitimate user who knows the original image can determine it.In addition, since many kinds of distorted images can be generated from one image, there is no need to worry about snooping even if the screen is recorded.

　In order to verify the effectiveness, 20 participants carried out 3 types of attacks (over-the-shoulder shooting, camera recording, screen recording) 300 times each as attackers using the existing method and EYEDi.As a result of examining the classification error rate of legitimate users and attackers, EYEDi was excellent.In particular, EYEDi proved to be effective in eliminating screenshot attackers, showing high protection against screen recordings of the most serious threat models.

　In the future, the company aims to create a system that is easy to use, such as making EYEDi multimodal and shortening the authentication time.

Paper information:[IEEE Access] EYEDi: Graphical Authentication Scheme of Estimating Your Encodable Distorted Images to Prevent Screenshot Attacks