Fighting Online Fraud with AI

As online travel bookings surge, so too does the complexity of safeguarding user data and preventing fraud. CNBC recently spoke with Siddhartha Choudhury, a Senior Product Manager at Booking.com, to gain insights into how the online travel giant leverages artificial intelligence (AI) to combat increasingly sophisticated security threats.

Booking a vacation online requires a significant leap of faith. Consumers entrust platforms with their sensitive financial information, personal details, and intricate travel plans. For a company like Booking.com, processing millions of transactions daily, maintaining that trust is paramount. AI is becoming the cornerstone of their strategy.

Countering Modern Online Fraud Requires AI Assistance

The sheer volume of data processed by Booking.com is staggering. The company’s security apparatus goes far beyond preventing stolen credit card use. It must detect and neutralize a wide array of malicious activities, including fake hotel reviews, deceptive marketing schemes, phishing attacks, and account takeovers.

“We employ AI across a wide spectrum of safety and fraud risk mitigation scenarios,” Choudhury explained. “This involves processing petabytes of data, encompassing events generated from applications, infrastructure, messages, and emails.”

To manage this data deluge, Booking.com utilizes a multi-faceted approach. “We leverage a combination of vendor-specific, built-in machine learning (ML) solutions alongside our in-house developed solutions to identify and counteract fraudulent attacks effectively,” Choudhury added. This hybrid approach allows Booking.com to take advantage of cutting-edge, off-the-shelf AI tools while also building proprietary systems tailored to the company’s specific threat landscape and business needs. This blend offers advantages in both speed of deployment and customization capabilities.

This strategy represents a deliberate approach to security, combining readily available software with custom-built AI to build a robust defense, protecting digital assets and user data. The key is identifying the correct tool for the job and integrating it seamlessly into the existing security infrastructure.

The Performance-Cost Equation

Operating a security system at this scale is undoubtedly challenging. Integrating disparate tools, both internal and external, represents a significant hurdle. But, the major concern is finding the proper balance between performance and cost.

As cyberattacks become increasingly sophisticated, defenses must evolve continuously to stay ahead of emerging threats. However, implementing more robust security measures comes at a higher price.

“The increasing sophistication and scale of cyber threats has upped the complexity,” Choudhury stated. He added, “The core decision is determining where to focus resources: on enhancing cost efficiency or improving performance?” This reflects a fundamental trade-off in cybersecurity investment – better security often translates to higher costs, and finding the optimal point on that curve is a critical strategic decision.

How AI Helps to Get Ahead of Online Fraud Threats

Proactive threat detection is crucial in modern cybersecurity. Instead of simply reacting to incidents after they occur, Booking.com is leveraging AI to identify and mitigate potential threats before they materialize. A major component of this strategy involved migrating systems to the cloud, and this enables smarter, faster tools.

Choudhury emphasized that their security experts work in tandem with AI-powered assistants. “Multiple AI assistants operate concurrently, augmenting the capabilities of security analysts, boosting their efficiency, and minimizing repetitive operational tasks,” he explained.

By providing human security analysts with AI-powered tools capable of sifting through massive datasets in real-time, they ensure experts can concentrate on the most critical and complex threats. These measures are complemented by robust monitoring systems to ensure the AI operates smoothly and accurately, guarding against biased or erroneous outputs.

Making Sure AI Plays Fair

When entrusting AI with critical security decisions, ensuring fairness and preventing bias is of utmost importance. Choudhury underscored the ethical considerations central to their AI strategy, focusing on several key principles:

• Fairness: The company conducts routine checks to detect and correct instances of AI bias, with particular attention to preventing the unfair targeting of specific individuals or groups as potential fraud risks. This involves continuously auditing the AI models and training data to identify and mitigate any unintended biases.

• Human Oversight: Human intervention is maintained for significant decisions. “Sufficient human input is involved to identify false positives,” Choudhury stated. A human, not just a program, has the final say. This safeguard helps to prevent the system from making decisions based on spurious correlations or flawed logic.

• Explainability: Opaque AI systems are avoided. It is important for the security team to understand the reasoning behind each decision to ensure accountability.

• Privacy: Booking.com’s AI systems are developed on a basis of “strong data protection, privacy, and compliance,” based on consumer consent.

What’s on the Horizon?

Choudhury believes the immediate future of AI in cybersecurity will focus on optimized integration. That means improving collaboration between existing tools.

“I anticipate a continuous development of new AI solutions, coupled with a focus on orchestrating these solutions to maximize departmental efficiency,” Choudhury predicted.

The overarching goal is to create an integrated security ecosystem where different components communicate and cooperate seamlessly. For Booking.com, the objectives are clear: “Driving innovation alongside guaranteeing reliability and cost efficiency will be the key focus.”

For travelers who are booking online, AI is providing security measures that help make the process safer.