Sessions » Identity applications of machine learning and data analytics

Identity applications of machine learning and data analytics

 

SILVER ROOM "THE LAB" - 206
Day 2 - May 1

 

Machine learning is an important frontier in next-generation identity solutions. Greater computing power and open source platforms are accelerating performance gains in a wide range of technologies, including facial recognition, behavioral biometrics, data analytics, and continuous authentication. The gains are so dramatic that they warrant a fresh look at policy, best practices, and technical standards – issues that this session will address head-on.

 

10:45
Introduction
Rob Atkinson, Founder and President, Information Technology & Innovation Foundation (ITIF), USA



10:50
Leapfrogging: The impact of machine learning on ID verification
Sunil Madhu, CEO, Socure, USA

Artificial Intelligence has the potential to go beyond the limitations of human intelligence and intuition to transform our approach fraud prevention and identity verification.  The application of artificial intelligence and machine learning to fraud prevention and identity verification, however, is poorly understood. Institutions and technology vendors are exploring the use of these analytical techniques using a variety of approaches and achieving a broad range of results. 

This session will explore best practices and future trends for the application of AI and machine learning to fraud prevention and ID verification.
•    What approaches, techniques and data elements have proven to provide the greatest level of insights in this space? 
•    What are some of the shortcomings of machine learning and artificial intelligence?
•    What does the future hold?
•    Revamped online identification verification review – more easily spot risks and detect fraudsters;
•    Value on analytics – not selling data;
•    Shifting from static identity to digital identity through machine learning and AI.

 

11:10
Bias in biometrics – How demographic factors impact security and usability
Brian W. Greene, Portfolio Manager, Border and Transportation Security, Defence Research and Development Canada (DRDC), Centre for Security Science (CSS), Department of National Defence, Canada

Demographic factors such as age, sex and ethnicity have long been known to impact biometric performance, but there have been few studies to properly understand the magnitude of these effects. Since 1995, the Government of Canada has used Gender-based Analysis Plus (GBA+) to examine how government programs affect individuals with different sex, gender and other relevant demographic factors. This was not applied to biometric systems, until 2016, when a new commitment to GBA+ led to the development of a report entitled “Bias in Biometric systems” that used data from operational systems to examine this issue. The preliminary results were compelling enough that Canada has initiated a new ISO technical report 22116 “Identifying and mitigating the differential impact of demographic factors in biometric systems”. Other countries have begun to contribute additional data and it is now apparent that age, sex and ethnicity all have significant impact on biometric performance. Women and children frequently experience higher failure to enrol rates, failure to acquire rates and false non-match rates than others. Even more disturbing, facial recognition, which is now used extensively for border control, has a significant security vulnerability where individuals from certain demographic groups can experience very high false match rates.

•    Demographic factors such as age, sex and ethnicity have a large impact on biometric performance;
•    Women, children and some ethnic groups experience difficulty using some biometric systems unless suitable measure to facilitate them are implemented;
•    Border security using facial recognition eGates is massively compromised for certain demographic groups.

 

11:30
Transparency, model governance and compliance
Ken Meiser, Chief Compliance and Consumer Support Officer at ID Analytics, USA

High-level review of an approach to model construction, testing and validation processes. The issues of model explainability transparency and communication to support end-user governance processes will also be discussed.

 

11:50
Questions and Answers


Share this page :
Follow us on :

Event Powered By


see more

Platinum Sponsor


see more

Gold Sponsors


see more

Silver Sponsors


see more

Bronze Sponsors


see more

Sponsor


see more

STRATEGIC PARTNERS


see more

Supporting Associations


see more

Premier Partners


see more

Media Partners


see more