MULTI-BIOMETRIC SYSTEMChris Chiffriller, Chris George, Gabriel Dos Santos, Subah Sachdeva

Biometrics• Over the past several years and with the growth of

technology, the field of biometrics has become very important in user identity and authentication.

• There is a growing need for these systems to be more secure and reliable

• The Fusion of multiple biometric traits can greatly improve both performance and reliability over approaches that only use a single biometric model.

Fusion• Fusion of multiple biometric traits can greatly improve

performance over unimodal approaches of user authentication.

• Fusion is placed into two broad categories, preclassification and post-classification. • Pre-classification fusion refers to combining information

prior to the application of any classifier or matching algorithm.

• In post-classification fusion, the information is combined after the decisions of the classifiers have been obtained

Fusion Levels• First one is sensor level fusion, by which raw data from multiple

sensors are fused into new raw data• Second one is feature level fusion, in which multiple features

acquired from feature extraction processes are fused into new feature vector

• Third level, score level fusion performs fusion of scores obtained from each matching process

• Fourth level is decision level fusion which carries out final decision based on the combination of individual decision results

Fusion Methods• Fusion Strategies Used

• Maximum • The Maximum fusion method gives the maximum score from different

modalities.• Median

• The Median fusion method gives the median score from different modalities

• Mean • The mean scores from different modalities are given by this method

• Minimum• Similar to the previous discussed methods, the Minimum fusion method

gives the minimum score from different modalities

Methodology• All data has been collected through the Pace University

Biometric (PBS) system and the Pace University Keystroke Biometric (PKBS) system.

• The data collection process was done in a closed system with a controlled environment. Sixteen students were given 10 online examinations with 10 questions for a duration of 20 minutes per exam.

• The mouse motion, keystroke and other biometrics sample data outputs, from each of the students and their examinations, were actively being accumulated and saved by the data collector as csv files in a database.

• This data was then taken and passed through a classifier, which produces matching score data by each of the biometric features

Anaconda & Fusion• Anaconda is a FREE enterprise-ready Python distribution

for data analytics, processing, and scientific computing. Anaconda comes with Python 2.7 or Python 3.4 and 100+ cross-platform tested and optimized Python packages. All of the usual Python ecosystem tools work with Anaconda.

• We used Python’s Scientific Libraries such as skearn and numpy to n provides insight on how the data is manipulated, classified and combined.

Using Anaconda • Step 1: Open the Anaconda Command Prompt

Using Anaconda • Step 2: Within the Command Prompt navigate to your

project source folder using the cd command (i.e. cd C:\Users\Chris\workspace\Fusion)

Using Anaconda• Step 12: Enter the name of the program followed by the

desired .csv data files (i.e. fusion.py click.csv scroll.csv)

Using Anaconda • Mix and match different features to compare the accuracy

of different features.

Future Goals• Find a way to ensure higher accuracy and lower ERR%

rates.• experiment using the scroll and the click features along

with the fusion methods• Being able to determine which data works better with

different fusion methods at different fusion levels