Robust continuous user authentication system using long short term memory network for healthcare

Tanveer, Anum, Lasebae, Aboubaker ORCID logoORCID: https://orcid.org/0000-0003-2312-9694, Ali, Kamran ORCID logoORCID: https://orcid.org/0000-0001-5301-9125, Alkhayyat, Ahmed, Ur-Rehman, Masood, Haq, Bushra and Naeem, Bushra (2021) Robust continuous user authentication system using long short term memory network for healthcare. Ur-Rehman, Masood and Zoha, Ahmed, eds. Body Area Networks. Smart IoT and Big Data for Intelligent Health: 16th EAI International Conference, BODYNETS 2021, Virtual Event, October 25-26, 2021, Proceedings. In: EAI BODYNETS 2021 - 16th EAI International Conference on Body Area Networks: Smart IoT and big data for intelligent health management, 25-26 Oct 2021, Glasgow, Great Britain (Online). pbk-ISBN 9783030955946. ISSN 1867-8211 [Conference or Workshop Item] (Accepted/In press)

[img] PDF - Final accepted version (with author's formatting)
Restricted to Repository staff and depositor only

Download (417kB)

Abstract

A traditional user authentication method comprises of username, passwords, tokens and PINs to validate the identity of user at initial login. However, a continuous monitoring method is needed for the security of critical healthcare systems which can authenticate user on each action performed on the system in order to ensure that only legitimate user i.e., a genuine patient or medical employee is accessing the data from user account. In this aspect, the perception of employing behavioural patterns of user as biometric credential to incessantly re-verifying the user’s identity is being investigated in this research work to make the healthcare database information more secure. The keystroke behavioural biometric data represents the organisation of events in such a manner which resembles a time-series data, therefore, the recurrent neural network is used to learn the hidden and unique features of users’ behaviour saved in timeseries. Two different architectures based on per-frame classification and integrated per frame-per sequence classification are employed to assess the system performance. The proposed novel integrated model combines the notion of authenticating user on each single action and on each sequence of actions. Therefore, firstly it gives no room to imposter users to perform any illicit activity as it authenticates user on each action and secondly it tends to include the advantage of hidden unique features related to specific user saved in a sequence of actions. Hence, it identifies the abnormal user behaviour more quickly in order to escalate the security, especially in healthcare sector to secure the confidential medical data.

Item Type: Conference or Workshop Item (Paper)
Research Areas: A. > School of Science and Technology > Computer Science
Item ID: 34586
Notes on copyright: This version of the contribution has been accepted for publication, after peer review (when applicable) but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: http://dx.doi.org/[insert DOI]. Use of this Accepted Version is subject to the publisher’s Accepted Manuscript terms of use https://www.springernature.com/gp/open-research/policies/accepted-manuscript-terms
Useful Links:
Depositing User: Kamran Ali
Date Deposited: 25 Jan 2022 14:33
Last Modified: 29 Nov 2022 17:45
URI: https://eprints.mdx.ac.uk/id/eprint/34586

Actions (login required)

View Item View Item

Statistics

Activity Overview
6 month trend
2Downloads
6 month trend
38Hits

Additional statistics are available via IRStats2.