An efficient privacy-preserving user authentication scheme using image processing and blockchain technologies. (19th May 2022)

• Record Type: Journal Article
• Authors: Ara, Anees; Sharma, Avinash; Yadav, Dharminder
• Is Part Of: Journal of discrete mathematical sciences & cryptography. Volume 25:Number 4(2022)
• Page Start: 1137
• Page End: 1155
• Publication Date: 2022-05-19
• Subjects: 68M25 -- 68P27 -- 68T07 -- 68T10 -- 68U10; Image processing -- Opencv -- CNN -- RCNN -- Imagechain -- Deepface -- Security -- User authentication; Computer science -- Mathematics -- Periodicals; Cryptography -- Periodicals; Computer science -- Mathematics; Cryptography; Periodicals; 004.0151
• DOI: 10.1080/09720529.2022.2075089
• Languages: English
• ISSNs: 0972-0529

Abstract

Abstract: User authentication is a measurement challenge for handheld devices and online accounts such as bank accounts, social media accounts etc. because illegal access results in money loss and user privacy. Individual devices, online financial services, and intelligent spaces are three significant areas of concern for customer authentication procedures. Three ways have been identified for authentication factors: i) knowledge-factor, ii) Inherence factor, and iii) possession-factor. This study investigates two-way user authentication through image processing. CNN, RCNN, and Deepface are deep learning algorithms used for image recognition. We used imagechain for image storage and Blockchain for personal information storage (mobile number) to secure the database. The database is stored on an Ethereum-based blockchain. After determining whether the image is fake or real, match the webcam image with the imagechain; if both images match, the one-time password is given to the user's cellphone number for login access. For image processing, Opencv is employed, and the Python library is used to execute machine and deep learning algorithms for user authentication. Test the proposed model on the 10 to 100 users for authentication. Accuracy of this experiment is 75.35, 76.33, 98.18 and cosine similarities of images are much better between images, but in case of fake image identification it achieved 97.35 % accuracy.