Computer Science > Information Theory
[Submitted on 13 Jul 2016 (v1), revised 29 Aug 2016 (this version, v3), latest version 20 Nov 2017 (v7)]
Title:The Partially Observable Hidden Markov Model with Application to Keystroke Biometrics
View PDFAbstract:This work introduces the partially observable hidden Markov model (POHMM), an extension of the hidden Markov Model (HMM) in which the hidden states are conditioned observed event types that form an independent Markov chain. Despite an explosion in the number of model parameters, parameter estimation can still be performed in time that scales linearly with the number of observations. Marginal distributions of the model act as a fallback mechanism when novel event types are encountered during likelihood estimation. A parameter smoothing technique accounts for missing data during parameter estimation and simultaneously reduces the degrees of freedom of the model to avoid overfitting. The structure of the POHMM is motivated by sequences that contain metadata which may partially reveal the underlying hidden state. Such a scenario is encountered in keystroke biometrics where the user is in either an active or passive state of typing, and the keyboard key names are event types that partially reveal the hidden state. The proposed model is shown to consistently outperform other anomaly detectors, including the standard HMM, in biometric identification and verification tasks and is generally preferred over the HMM in a Monte Carlo goodness of fit test.
Submission history
From: John V Monaco [view email][v1] Wed, 13 Jul 2016 18:30:33 UTC (2,541 KB)
[v2] Tue, 23 Aug 2016 14:03:42 UTC (3,156 KB)
[v3] Mon, 29 Aug 2016 15:22:22 UTC (3,156 KB)
[v4] Wed, 21 Dec 2016 20:54:21 UTC (3,105 KB)
[v5] Tue, 11 Apr 2017 17:22:17 UTC (1,907 KB)
[v6] Fri, 17 Nov 2017 14:48:33 UTC (1,911 KB)
[v7] Mon, 20 Nov 2017 16:02:39 UTC (1,919 KB)
Current browse context:
cs.IT
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.