MLFraud (MeetUp)

ML Breakthroughs in Financial Security

26

November 2024

18:00-20:00

Petach-Tikva

Intuit offices

Ha-Psagot Street 7-9, Bldg B, 10th Fl

About The Conference

We invite you to our upcoming joint event with Intuit, MLFraud, to discover how machine learning impacts financial fraud prevention.

Our expert speakers will showcase real-world implementations, from new credit risk models to graph-based fraud ring detection techniques. We will dive into the technical challenges and innovative solutions influencing fraud prevention.

Don't miss this opportunity to explore the latest ML architectures, discuss implementation strategies, and network with the top experts in financial tech.

Register now to gain insights that will elevate your ML and data engineering skills in the fight against financial fraud!

Who is it for?

It is ideal for advanced and senior AI researchers, data scientists, data engineers, ML engineers, VP R&D, and academics interested in exploring AI technology in financial fraud.

The Event Details

  • 26.11.2024

  • 18:00-20:00

  • Ha-Psagot Street 7-9, Bldg B, 10th Fl, Petach Tikva, Israel

Speakers

Racheli Lazar

Staff Data Scientist

Intuit

Lotem Even Zahav

Data Lead

Fido

Gal Benor

Machine Learning Scientist

Paypal

Agenda

18:00-18:30

Gathering and mingling

× ×

18:30-19:00

Utilizing User Activity events for User Segmentation

× ×

In-product User activity events can be a valuable signal for user segmentation and user clusterin – which can be used both for personalization purposes, but also for fraud detection. In the model we'll present, we evaluate the similarity between users by leveraging their in-product activity events. By considering each user as a document composed of his activity events, we build the user embedding using good old TF-IDF, but with a twist.
Prior to calculating the similarities between users, we significantly reduce the huge initial number of pairs of users by employing Locality Sensitive Hashing (LSH) techniques. The problem is then represented as an undirected graph, with users as nodes and the similarities between them as weighted edges, and community detection methods are applied to reveal clusters of users with similar in-product behavior.
To improve the stability of the resulting clusters and allow for a methodological way to evaluate various features, we leverage clustering ensemble methods.
By incorporating various expert-knowledge labels and meta-features on the entire users population, we show how resulting clusters are correlated with suspicious activity, and how it can help eliminate abusive usage in the product.

Racheli Lazar

Staff Data Scientist at Intuit

19:00-19:30

Leveraging ML Tools for Advanced Fraud Detection

× ×

How do you offer instant credit to millions of unbanked people while staying one step ahead of increasingly sophisticated fraud? At Fido, we’re tackling this challenge head-on, using cutting-edge machine learning and advanced risk models to provide financial access to individuals who have no traditional credit history. By leveraging real-time data and predictive analytics, we make fast, accurate decisions, all while ensuring our systems can scale securely as we expand across Sub-Saharan Africa.

In this talk, I’ll dive into our machine learning-driven strategies for combating fraud and managing risk. I'll highlight our geo-hashing model, which encodes geographic data to efficiently detect location-based anomalies, and our innovative "fake selfie" detection model, which identifies fraudulent submissions where users attempt to pass off pictures of pictures as legitimate IDs. Additionally, we’ll explore other fraud detection techniques, offering insights into the challenges we face and the advanced solutions we’re implementing to outsmart evolving threats.

Lotem Even Zahav

Data Lead at Fido

19:30-20:00

Opening the Black Box : Harnessing XAI for Revealing User Patterns in Fraud Detection

× ×

A wise man once told me, “It is not only the WHAT that matters; rather, it is the WHY.” In an era where fraud is becoming increasingly sophisticated, the need for transparency in machine learning models is more critical than ever. This session will explore how explainable AI (XAI) can demystify the decision-making processes behind fraud detection algorithms. We will discuss the importance of understanding user behavior patterns and how XAI can provide actionable insights, enabling detection of fraud more effectively and explaining these decisions for better outcomes. Discover how XAI transforms the landscape of fraud detection, enhancing trust and accountability while maximizing the value of each prediction.

Gal Benor

Machine Learning Scientist at Paypal