Interview with Kevin Lee, VP of Trust & Safety at Sift

Shauli Zacks Shauli Zacks

SafetyDetectives recently sat down with Kevin Lee, the VP of Trust & Safety at Sift, for an insightful conversation about the evolving landscape of online fraud. Delving into which industries are most at risk, Lee illuminated how Sift employs AI and machine learning as powerful tools against cybercrime. He further discussed the unique challenges presented by cross-border transactions and highlighted the intricacies of international fraud patterns. This comprehensive discussion sheds light on the multifaceted nature of online security in today’s digital age.

Hi Kevin, Thank you for your time. Can you tell me about your journey and current role at Sift?

I’ve always been passionate about making the Internet a safer and more trusted place. Before joining Sift, I led the Global Spam Operations team at Facebook (now Meta), which was tasked with keeping both Instagram’s and Facebook’s news feeds free of various kinds of spam. Prior to that, I was the Head of Risk at Square, where I was working to combat and prevent payment fraud. I started my career at Google, also in a role focused on fighting online abuse.  At each of those companies, I built, scaled, and managed various abuse teams, and I’ve taken all the lessons learned in those roles and applied them to my work at Sift. 

I’ve worn a lot of different hats at the company in my seven years at Sift, but my current role is VP of Trust & Safety. I head up a team of talented Trust & Safety Architects who are working to stay ahead of fraud trends by doing on-the-ground research into new and evolving fraud trends, as well as analyzing fraud patterns and behaviors on behalf of customers. We then use those insights to develop industry best practices and standards, and to inform our product strategy – for us, this means designing and operationalizing systems for detection and policy enforcement at scale. I also spend a lot of time interfacing with our customers to understand their most pressing concerns as the fraud landscape continues to evolve, and help them balance fraud prevention against user friction. 

Can you tell me about Sift and what are the main services?

Sift’s Digital Trust & Safety platform was built to give companies the most holistic solution to manage a wide variety of fraud types. We pride ourselves on the level of control and transparency we offer businesses, as well as the size and diversity of our global network. We work with over 34,000 sites and apps across a range of industries, and process over 1 trillion events annually. Sift offers customers global machine learning models that are fueled by a shared intelligence data network, as well as custom models that are tailored to a specific business, so companies can enable proactive decisions that stop fraud before it happens. In helping businesses balance growth against risk, we ultimately help accelerate revenue growth and reduce financial losses.

What industries or sectors are most vulnerable to online fraud?

With more consumer activity happening online, almost all companies today have some sort of digital presence. But this transition to digital platforms also means that almost all businesses are coming face-to-face with online fraud. Cybercriminals have had their sights set on e-commerce for a long time, and any merchant with an online presence faces some risk; Sift network data shows that attempted payment fraud in fintech jumped 13% between 2021 and 2022. This year, card-not-present (CNP) fraud is projected to account for almost $10 billion in losses, making up 73% of all payment fraud. And with the uptick in increasingly accessible fraud forums on the deep and dark web (what we call “the democratization of fraud,” or Fraud-as-a-Service), businesses are seeing a relentless influx of attacks.

How does Sift leverage AI and machine learning to tackle online fraud effectively?

AI is rapidly changing the fraud landscape as we know it –– bad actors are using AI to conduct more sophisticated scams and can quickly scale fraud campaigns thanks to automation. Sift research reveals the frequency of fraud is on the rise; a staggering 68% of consumers reported an increase in the frequency of spam and scams in the six months since tools like ChatGPT were released.

In order to fight AI-enhanced fraud, we have to use AI to get ahead of it. Machine learning plays a massive role in fraud detection and prevention and can empower companies to even the playing field. For example, ML can analyze large volumes of data in real-time to swiftly identify patterns, anomalies, and trends indicative of fraudulent behavior. These capabilities can be amplified by global networks of fraud signals, which in turn enables organizations to detect fraud more accurately, minimizing false positives and enhancing overall detection capabilities.

Sift’s real-time machine learning also reduces friction for legitimate customers by adapting security measures based on the level of risk associated with each transaction or user interaction using an approach called ‘dynamic friction.’ Being proactive in preventing fraud is a key component to protecting businesses and their customers.

Considering the global nature of online fraud, how does Sift address challenges related to cross-border transactions and international fraud patterns?

Digital native companies are increasingly border-agnostic, with transactions that occur wherever customers or partners might live, so Sift’s approach has always used a global lens. Most recently, we announced a significant expansion of our worldwide Sift Partner Program to extend our fraud prevention platform to a network of businesses worldwide, and across key industry verticals. The expanded program supports business models and partner success across multiple partner types, including Solutions Partners, Payment Service Providers (PSPs), Payment Orchestration Platforms, and Managed Services Providers (MSPs), in addition to Technology integration partners via the Sift Connect library of open APIs.

Beyond that, our (unrivaled) network of data is global by default, and represents shared signals from Sift customers across industries, regions, and fraud vectors that feed into our machine learning models. Across 34k+ sites and apps worldwide, Sift ingests more than one trillion events per year, which are categorized into 16k different signals of potentially fraudulent activity. This means our network is able to detect new attack patterns within 250 milliseconds, providing customers across the Sift network with near-instant protection. When fraudsters launch an attack on one Sift customer, that data is automatically fed into our machine learning models and blocked from impacting other customers in the network.

As fraudsters constantly adapt their tactics, how can companies continuously adapt and stay proactive in identifying new fraud patterns?

As AI-enabled fraud becomes more common, businesses must respond. By leveraging advanced analytics, machine learning, and dynamic friction, companies can enhance fraud detection capabilities, enable better customer experiences, and proactively combat evolving fraudulent activities. But it’s hard for businesses to achieve this on their own, the ability to respond quickly to fraud threats and scale these efforts is more important than ever  –– it would be impossible for most businesses to build the technology and tools in-house to effectively fight fraud. 

Whether directly working together as a criminal enterprise or indirectly working together through sharing of information on the web, cybercriminals collaborate in order to steal from consumers and businesses. It takes a network to fight a network, and this is why Sift has invested in ensuring that our capabilities –– both human and technological  –– help to connect fraud fighters across the globe. 

About the Author

About the Author

Shauli Zacks is a tech enthusiast who has reviewed and compared hundreds of programs in multiple niches, including cybersecurity, office and productivity tools, and parental control apps. He enjoys researching and understanding what features are important to the people using these tools.