Mouse Movement Project

1
Mouse Movement Project

• Customer Larry Immohr
• Professor Dr. Charles Tappert
• Team
• Shinese Noble
• Anil Ramapanicker
• Pranav Shah
• Adam Weiss

2
Agenda

• Brief description of project
• Project Requirements
• Meetings
• Design Decisions
• Components
• Testing Strategy
• Challenges
• Questions

3
Description of Project

• Creation of a pattern recognition system for
  Mouse Movement Biometrics
• Identify computer users thru their individual
  mouse movements
• Provides a feasibility study on whether this is a
  relevant way to track computer users behavior and
  identify them.

4
Description of Project

• Enough individuality in a users mouse movements
  to identify them
• Arc
• Speed
• Acceleration / Deceleration
• Clicking
• Collect Data ? Identify Features ? Classify User

5
Mouse Movement Biometric System- High-Level View
6
Project Requirements

• Capture data of individual mouse user
• Mouse Movement
• Mouse Click
• Store data
• Perform calculations to quantify mouse movements
• Utilize data to identify user

7
Meetings

• Team met via phone conference every Monday
• Constant communication via e-mail
• Meeting with client via phone conference every
  Tuesday
• Communication via e-mail
• Sharing documentation

8
Design Decisions

• Modular Format
• Runs in Background
• Can be layered with any application
• Utilizes an enrollment program to get
  fingerprint of user
• Focused on a limited number of features due to
  time and resource constraints
• Additional requirements can be built in as
  project continues

9
Components

• 3 modules of the program
• Data Capture
• Feature Extraction
• Classification

10
Design Decisions 3 Programs
11
Data Capture Architectural View
12
Data Capture - Enrollment
13
Data Capture - Data
Mouse User Action
Event Time in Milliseconds
X Coordinate
Y Coordinate
14
Feature Extraction

• Reads the raw data file
• Parses data into mouse curves and mouse clicks
• Compute individual curve and click measurements
• Creates a mouse profile of user
• Creates mouse profile measurements

15
Feature Extraction View
16
Feature Extraction - Data
17
Classification

• Takes the feature vectors as the input
• Normalizes the data
• Uses K-Nearest Neighbor algorithm for a test case
• Does a leave one out method for cross validation
  between many cases
• Prints out the matching cases
• Analyze the cross validation results and prints
  out the success statistics

18
Testing Strategy

• Multiple releases
• Testing amongst team for bugs
• Delivered to client after team testing
• Repeated for each release
• For program data, all members input 5 samples of
  data

19
Challenges

• Establishment of clear goals
• Change in scope of project
• New project many unknowns
• How to utilize enrollment program

20
Questions?