A Recommender-Based Graphical Authentication System

Abstract

Graphical Authentication System (GAS), which requires the use of visual images or drawings as passwords, is now being adopted by many designers of information systems. However, most of the existing graphical authentication systems are not quite efficient because they do not consider operational and environmental factors. In addition, GAS is generally faced with the problems of shoulder surfing and visual dictionary attacks. In this study, an improved Recommendation technique for Graphical Authentication System (RGAS), which combines recognition and pure recall-based techniques, will be developed to address these problems. During authentication, Template matching analysis was used for representation and comparison of signs that are presented on a size GxG grid cells by the users. The design was implemented using JAVA scripts and WAMP client-server environment. The evaluation experiment was further conducted in the e-library unit of Federal College of Education, Abeokuta, with two hundred computer systems. The RGAS with other two existing schemes will be incorporated into the College library system and two hundred purposively selected participants were used and monitored for three months. The usability and security of RGAS were compared with A-Free Draw Graphical Password System (AFDGPS) and Hybrid Graphical Password Based System (HGPBS). The considered usability factors during evaluation are Efficiency (E), Learnability (L), Error rate (Er), Memorability (M) and Satisfaction (S). The performance of RGAS were evaluated using Accuracy Rate (ACR), Adaptation Rate (ADR), Success Rate (SR) and Mean Completion Time (MCT). Based on these usability and security evaluations, the result of the comparison of RGAS together with the existing AFDGPS and HGPBS are carried out to determine whether RGAS will yield better usability and security performance and that it will prevent all forms of common authentication attacks which could also be adopted by designers to enhance system security.

Keywords: Graphical password, recognition-based algorithm, cued recall-based algorithm, pure recall-based algorithm, RGAS, AFDGPS, HGPBS.

DOI: 10.7176/JIEA/14-2-01

Publication date:March 31^st 2024