Battling Multi-Billion Dollar Online Fraud Industry with AI and Machine Learning AI Expert Anjanava Biswas on how ML and AI are used to tackle online fraud and abuse

The digital age has brought unparalleled benefits, reshaping global communication and business operations. Yet, this interconnectivity introduces new risks as cybercriminals exploit online platforms. From phishing scams that harvest personal information to elaborate imposter schemes that mimic legitimate institutions, the methods of deception are becoming increasingly intricate, with the scale of online fraud reaching billions of dollars in losses each year.

So far, in 2022 alone, reported fraud data amounted to an alarming $8.8 billion. Investment and imposter scams topped the list, and eCommerce fraud also surged to $41 billion. Alarmingly, a projected increase of $130 billion in fraud losses by 2023 signals a dire need for effective countermeasures.

Recognizing the pressing need to protect sensitive data amidst escalating cyber threats, the technology giants, such as Azure, Google, and Amazon Web Services (AWS), have diligently fortified their defenses, developing specific fraud detection mechanisms to address their clients' specific online safety concerns.

For AWS specifically, Anjanava Biswas, a Senior AI Specialist Solutions Architect, stands at the forefront of this effort, advocating robust cybersecurity protocols and crafting effective countermeasures to address cybercriminal activities using AI models.

Biswas spearheads the deployment of AI-driven platforms, which are powerful tools for the early identification and interception of fraudulent account creations on digital services. As Biswas asserts, this proactive approach is a formidable barrier against fraud and abuse.

Tackling the Complexities of Online Fraud

Tackling online fraud presents a multifaceted challenge, with fraudsters continually refining their tactics, also known as "attack vectors." Conventional methods, reliant on fixed algorithms and historical data, fail to catch novel or rapidly mutating fraud strategies. This limitation exposes businesses and consumers to the dual risks of financial loss and personal data breaches.

Biswas highlights the daunting task ahead, stating, "We're combating a relentless enemy—one that's as resourceful as it is ruthless." The task is complicated by balancing security with user experience; overly stringent measures can impede legitimate transactions, while lenient ones invite exploitation. This delicate equilibrium is where AI and Machine Learning (ML) shine, providing a nuanced, real-time response to emerging threats.

"Platform abuse and fraud prevention largely remain reactive and is achieved by studying a user's profile behavior and transaction history after they sign up. This approach is often manual, time-consuming, and expensive. Early detection and prevention of fraudulent account sign-ups on online platforms using artificial intelligence (AI) is an effective defense mechanism for combating fraud and abuse," Biswas states in his 2021 technical paper.

With over fourteen years of experience in global supply chain, manufacturing, and online retail channels, Biswas brings a deep understanding of the dynamic nature of online fraud. This expertise informs the advanced solutions he develops, utilizing AI to counter these threats effectively.

An AI Based Fraud Prevention System

Central to the fight against online fraud, Biswas has been instrumental in the innovations and effective usage of Amazon Fraud Detector, an AI-based fraud prevention system. This fully managed machine learning (ML) service leverages over two decades of Amazon's deep fraud detection expertise, honed through its vast eCommerce ecosystem, to address a broad spectrum of fraudulent activities, from new account creation to checkout and promotional fraud, account takeovers, and credit card fraud. It uses state-of-the-art ML algorithms and NLP (natural language processing) to detect and adapt to the nuances of fraudulent online behavior. It covers various fraud scenarios, from new account creation to checkout, account takeovers, and promotional fraud.

According to Biswas, the operational elegance of the Amazon Fraud Detector lies in its seamless integration during the customer sign-up process, as explained in his 2021 technical paper. Its design is a supervised ML model along with NLP that works with Amazon Cognito's user sign-up workflow for online platforms and mobile applications. Additionally, Biswas' research resulted in a solution that includes Graph databases to tackle email aliasing, a common technique spammers use. This integration enables real-time fraud prevention for web and mobile applications, a feature crucial for the ever-expanding digital ecosystem.

When a new user registers, the system immediately assesses the risk. Low-risk cases proceed without interruption, while high-risk evaluations prompt additional verification to confirm identity. This strategy optimizes security and enhances the user experience by reducing friction for genuine users.

Moreover, as it employs AI algorithms, it analyzes historical fraud trends. This analysis goes beyond cataloging past fraud—it deciphers the subtle signs that often signal impending fraudulent actions, sharpening its predictive accuracy. Biswas has streamlined the complexity, making the tool accessible to businesses that lack extensive ML expertise, thus offering a user-friendly yet powerful fraud detection platform.

Biswas' Solution in Action

Amazon Fraud Detector has demonstrated versatility and robustness across diverse industries, proving a vital asset in the arsenal against fraud. Biswas elaborates that UK-based SLA Digital, a company specializing in digital services and telecommunications solutions, has utilized its system to safeguard against fraudulent transactions that could compromise revenue streams and customer trust. Similarly, in the travel industry, FlightHub Group has applied this innovative AI solution to protect against the unique threats that emerge in the high-volume transaction environment characteristic of travel bookings and accommodations.

At the same time, companies like GoDaddy, Qantas Loyalty, and even Standard Bank Insurance have harnessed the state-of-the-art capabilities of this Amazon Fraud Detector-powered solution to secure their vast ecosystems. These platforms, as prime targets for fraudsters due to the sheer number of transactions and sensitive customer data they handle, have benefitted from AI's ability to learn and adapt to the specific types of fraud prevalent in their respective fields.

Biswas presenting Amazon Fraud Detector at AWS Mega Modernization Week 2021

Biswas's expertise and the impact of his research are not limited to his technical paper. His presentation in the 2022 AWS AI & Machine Learning Modernization Week allowed him to share deep insights into the workings of Amazon Fraud Detector, showcasing its capabilities to industry professionals and the AI community. Moreover, during the prestigious AWS re:Invent 2020 conference, his solution was highlighted through a comprehensive video demonstration, offering a tangible showcase of its functionality and effectiveness.

Continuing the Fight Against Fraud

Biswas acknowledges that no system is immune to potential weaknesses in our era of rapid data exchange. Despite the strength of Amazon Fraud Detector, the volatile nature of fraud means that systems must remain agile. He affirms that the commitment to security begins with deployment and continues with ongoing refinement, anticipating the development of even more advanced AI algorithms in the future.