Mouse Movement Biometrics

1
Mouse Movement Biometrics

• Fall 2007 Capstone -Team Members
• Rafael Diaz
• Michael Lampe
• Nkem Ajufor
• Mohammed Islam
• Antony Amalraj

2
Mouse Movement Biometrics -Agenda of Final
Presentation

• Brief Scope of the project
• Project Requirements and Specification
• Design Decisions
• Objectives
• Demonstration of MMSystem
• Testing Strategy
• Meetings Format
• Challenges
• Wrap up/Summary of Accomplishments
• Recommendations
• Questions

3
Mouse Movement Biometrics Brief Scope of the
project

• This semester's project had two primary focuses.
• First, we became familiar with the system and
  collected as much additional data as possible,
  including data from each team member and third
  party data
• Second, and most importantly, we formatted the
  feature-vector data for ease of processing by
  other back-end teams, by normalizing
  feature-vector data.

4
Mouse Movement Biometrics Project Requirements

• Capture data of individual mouse user (a total of
  50 data files)
• Mouse Movement
• Mouse Click
• Generate corresponding feature data in normalized
  feature format for the backend teams
• Perform calculations to quantify mouse movements
• Obtain recognition accuracy (just first-choice
  nearest neighbor) using the leave-one-out
  procedure on the 50 data files.

5
Mouse Movement as a Biometric Measurement -
Specifications

• Mouse Movement data was captured through the
  enrollment process and the creation of a user
  profile
• The intent of data capturing is to identify the
  user based on the stored data and the data that
  was recently captured
• This method of identification, in which the data
  recently captured is compared to the information
  on the database, is known as a One to Many
  comparison
• Throughout this phase of the project the data
  captured and used was validated with both the K-
  Nearest Neighbor and Leave One Out methods
• Below are some of the focus points in this Mouse
  Movement Study
• Obtain data while user clicking buttons or
  enrolling user info
• Capture the data in a CSV format for
  normalization purposes
• Generate feature extraction data from feature
  extractor module.
• Classify user and possible identification using
  classifier
• Send a set of normalized data to backend teams
  and Generate
• success statistics

6
Mouse Movement Biometrics Design decisions
7
Mouse Movement Biometrics Objectives

•  
• We reported a total of 205 data files - including
  the data generated by 3rd parties
• Generated normalized feature vector data files
  and passed it on to the backend teams (Team 5 and
  6)
• Obtained recognition accuracy (first-choice
  nearest neighbor 80) using the leave-one-out
  procedure using 35 data files.
• Obtained results from KNN method using Classifier
  Module.

8
Mouse Movement Biometrics - Objectives contd

• Generated Data at weekly intervals - 205 files
  total, including 3rd party data
• Data from more subjects
• Data from random button sequences
• Enhanced mmsystem module has been developed with
  rich GUI features for the future users.
• It also will generate an additional file called
  profile.txt along with Raw data files.
• This Profile.txt file will be used as an input
  for both feature extraction and classifier
  module.
• The team created a website to ensure all our
  documents, course software will be uploaded in a
  centralized location.

9
Mouse Movement Biometrics - Objectives contd

• Enhancement of the existing front-end
  registration process that captures pertinent
  information regarding the user
• User Name
• Output File Name
• Gender gtgt Male or Female
• Age
• Right- handed or Left- handed
• Type of Mouse

10
Mouse Movement Biometrics - User Input GUI
Input Dialog Box 1 Enter User Name
Input Dialog Box 4 Select your age ( 18-50 or
N/A )
Input Dialog Box 2 Enter File Name
Input Dialog Box 3 Select your Gender ( Male /
Female )
11
Mouse Movement Biometrics - User Input Continued
Input Dialog Box 5 Select your hand used (
Right-handed / Left Handed )
Input Dialog Box 6 Select type of mouse (
Optical Mouse / Serial Mouse USB Mouse /
Wireless Mouse )
Input Dialog Box 7 Select type of Test Screen (
Fixed 25 button sequence, Tic-Tac-Toe Game, or
Blank Screen )
12
Mouse Movement Biometrics - User Input GUI to
be continued
13
Mouse Movement Biometrics-Normalized Feature
Vector Report
14
Mouse Movement Biometrics Demonstration of
Enhanced System

• Demonstration of the enhanced mouse movement (old
  mmsystem and new mmsystem) provide
  recommendations
• Overview of the Technical paper

15
(No Transcript)
16
Mouse Movement Biometrics Testing strategy

• Validation of the new code introduced to correct
  and address any bugs identified in the testing
  window
• Corrections to program/bugs done by team members
  after response/comments received from team and
  volunteers that ran the application
• For program data, all members input 5 samples of
  data and data was validated through the
  classifier program

17
Mouse Movement Biometrics - Meeting Format

• Team 1 met twice a week via a conference bridge
  Tuesdays and Fridays
• Tuesdays meeting was focused on the team and the
  overall status of the project
• Fridays meeting was focused on questions that
  were presented to the client
• All conference calls lasted 1 hour in duration
• Communication via e-mail was also used and all
  involved parties were copied on the e-mails.

18
Mouse Movement Biometrics Wrap up/Summary of
Accomplishments

• Captured raw data in a CSV format for
  normalization and experiments
• Generated Feature vector extraction data and
  Normalized Feature Vector
• Generated Data in Mushroom data format for data
  mining project
• Classified the users by KNN method and Leave One
  Out method
• Generated Classified output data and Success
  statistics Report
• Enhanced software modules to incorporate the GUI
  changes
• Generated the data in the required format
• Created Mouse Movement Biometrics Technical
  Paper
• Created a User Manual to document use of the
  software
• Created website to store the current application
  modules, tested results
• Created training videos for the three
  applications in order to assist users in learning
  the system.
• Uploaded the Technical Paper, User Manual as well
  as the mid term and final presentations on the
  website

19
Mouse Movement Biometrics Challenges

• During the initial enrollment process questions
  surrounding the application and how to access and
  run the application existed
• Difficulties in understanding the normalization
  process and using only two values (0 and 1)
• Getting the enhancements made for the existing
  MMSystems GUI to work in a single display window

20
Mouse Movement Biometrics - Recommendations

• Further enhancements to the data Capture module.
• Work was started on adding new data fields to
  the Feature Extraction and Classifier modules but
  will need to be continued by succeeding teams.
• While 100 accuracy is not probable, it seems
  more experiments need to be performed to see if
  there is a more consistent accuracy rate over
  time and from more generated data.
• Subsequent teams should focus on developing the
  Data Capture GUI to randomize the buttons to
  provide more varied data

21
Mouse Movement Biometrics Recommendations contd

• Subsequent teams can add more user
  characteristics to classify the user
• Also, Subsequent teams can add more
  characteristics of the mouse such as right click
  or track wheel use
• Finally, it would be optimal if the system was
  developed to be used online with a database
  backend.
• This would allow for more data to be generated
  from a larger pool of users for further analysis
  and research .

22
Questions/ Comments

• http//utopia.csis.pace.edu/cs691/2007-2008/team1/
  default.htm