Project Description

• Motivation: Users often forget their passwords or use simple and un-secure passwords. Machine Learning algorithms are capable of continuously verifying users based on their activities (e.g. mouse or keyboard usage) after initial authentication. By combining different methods for verifying the user’s identity, the overall security for the user can be increased.

• Approach: The project will develop state-of-the-art technology for verifying users’ identities based on their computer usage patterns, and combine this information with traditional (e.g. password-based) verification in order to increase privacy and security.

• Benefit: The verification system will improve both security and usability, and reduce the potential damage inflicted by identity theft and social engineering.

Objectives

• Verify a user’s identity based on his or her activities when accessing his or her account.

• Combine this innovative verification method with others in order to improve privacy and security.

• Develop an advanced prototype for demonstrating the new technology and facilitating empirical evaluation.

• Evaluate the prototype and identify the best parameters for calibrating the learning and verification algorithms, in order to maintain a high level of verification accuracy.

• Derive guidelines regarding the usability of such an approach, in order to increase user acceptance.

• Realize a solution that will address privacy concerns.

• Identify and describe use-cases for activity-based verification

Time Frame:
Phase 1: 07/2008 - 03/2009;
Phase 2: 04/2009 - 06/2010

T-labs Team Members:Sebastian Möller, Niklas Kirschnick

Partners: Deutsche Telekom Laboratories at Ben-Gurion University, DAI-Labor/Technische Universität Berlin

Funding by:Deutsche Telekom Laboratories