PayPal’s CISO on how generative AI can improve cybersecurity

VentureBeat just lately sat down (nearly) with Assaf Keren, CISO and VP of enterprise cyber safety at PayPal. Keren’s profession at PayPal spans eight years and features a broad base of expertise constructing safety merchandise, managing safety infrastructure and creating and main safety companies.

Earlier than PayPal, Keren based two cybersecurity startups, served as head of safety product administration at Verint and was CISO for the Israeli e-Authorities Division (Ministry of Finance).

PayPal’s AI-driven cybersecurity technique is paying off

For his or her newest quarter (FY Q2 2023), PayPal reported $7.3 billion in income, up 7% year-over-year, or 8% on a overseas currency-neutral (FXN) foundation. Stable good points in transaction development of 5% and a 37% improve in value-added companies income contributed to a robust quarter. PayPal ended the calendar 12 months 2022 with $27.5 billion in income and has a trailing twelve-month (TTM) income run charge of $28.5B.

PayPal’s complete fee quantity grew 11%, supported by 10% transaction development, whereas tight expense management drove an 11% decline in non-transaction prices. Regardless of a $1.2 billion influence from loans originating for the pending BNPL sale, PayPal generated sturdy free money circulation and returned $1.5 billion to shareholders. General, PayPal confirmed wholesome quantity and consumer engagement development, considerably improved profitability, and progress on strategic initiatives.

Recognized for the depth of their experience utilizing AI, machine studying (ML) and pace of coaching and deploying deep studying fashions to harden cybersecurity and cut back fraud, PayPal’s 11% discount in losses within the newest quarter amidst rising volumes point out AI is delivering worth at scale. Reaching a double-digit discount in losses by way of improved danger administration signifies that the corporate’s AI-powered cybersecurity defenses, fraud detection fashions and AI-powered danger administration are decreasing transaction losses.

The next is an excerpt of VentureBeat’s interview with Keren.

VB: How can generative AI bolster safety with out negatively impacting the client expertise?

Keren: Whereas gen AI could also be utilized by attackers for malicious functions comparable to producing false identities or creating malware variants that evade conventional safety measures, it would additionally empower security-forward organizations like PayPal to discover gen AI-driven protection mechanisms, comparable to next-generation automated menace detection techniques and response capabilities. We have now all the time put safety on the middle of what we do and construct merchandise with safety in thoughts by design. With gen AI, there might be alternatives to reinforce our capabilities, take away friction and drive larger buyer worth.

VB: What are the dangers related to utilizing gen AI, and the way can these be mitigated for platforms dealing with a considerable amount of non-public information?

Keren: Gen AI, like several new expertise, have to be fastidiously evaluated and deployed with security and duty as high priorities. Vital components like information high quality, mental property, safety, privateness and compliance are key issues. To mitigate dangers, it’s vital for expertise leaders to look at these dimensions scrupulously earlier than deploying gen AI to guard prospects and protect their belief. Altering the methods we take into consideration testing and offensive safety to incorporate assault patterns comparable to immediate injection, checking for bias and figuring out the place the fashions hallucinate will enable gen AI to be deployed in a approach that’s helpful to prospects.

VB: Can gen AI assist with fraud detection? In that case, how will you validate its accuracy?

Keren: Sure, and in a way, it already is. PayPal has been an early adopter of AI, and we’ve been constructing our AI capabilities and experience for over a decade. We’ve been using transformer-based deep studying for years, which is the important thing expertise behind large language models (LLMs) like ChatGPT. Immediately, we use AI and ML throughout broad domains in our enterprise, together with fraud discount, buyer safety, personalised companies, danger administration and world commerce empowerment.

VB: Are you able to share the outcomes PayPal is attaining utilizing AI for fraud detection?

Keren: It’s led to large influence: from 2019 to 2022, at a time when our annual fee volumes practically doubled from $712 billion {dollars} to $1.36 trillion {dollars}, we reduce our loss charge by practically half, partly because of our advances in AI algorithms and expertise.

Immediately, with our advances in AI, we’re capable of quickly adapt to altering fraud patterns to guard our prospects. PayPal’s deep learning fashions will be skilled and pushed to manufacturing inside 2 to three weeks and even faster for less complicated algorithms. This enables us to coach fashions with the most recent manufacturing information, incorporate new fraud patterns and combine suggestions from inside brokers and PayPal prospects.

VB: How does the quickly altering monetary expertise panorama affect gen AI technique adoption?

Keren: As an organization with near 430 million lively accounts and 35 million retailers on our two-sided community, we see large potential within the energy of AI to create the subsequent technology of checkout and commerce. Immediately, PayPal holds over 200 petabytes of funds information — information that’s key to our aggressive benefit that holds large insights and potential to energy higher commerce experiences for shoppers and retailers.

We’re actively assessing gen AI and its potential influence on our enterprise and the aggressive panorama, and we’re enabling inside groups to think about the place gen AI can improve effectivity, improve our safety, enhance the client expertise, shield our prospects and develop our enterprise.