Inside the Briefcase

Ironclad SaaS Security for Cloud-Forward Enterprises

Ironclad SaaS Security for Cloud-Forward Enterprises

The 2015 Anthem data breach was the result of...

The Key Benefits of Using Social Media for Business

The Key Benefits of Using Social Media for Business

Worldwide, there are more than 2.6 billion social media...

Infographic: The Three Pillars of Digital Identity: Trust, Consent, Knowledge

Infographic: The Three Pillars of Digital Identity: Trust, Consent, Knowledge

8,434 adults were surveyed to gauge consumer awareness of...

FICO Scales with Oracle Cloud

FICO Scales with Oracle Cloud

Doug Clare, Vice President at FICO, describes how Oracle...

Is Your Enterprise IT the Best It Can Be?

Is Your Enterprise IT the Best It Can Be?

Enterprise IT is a driver of the global economy....

How Machine Learning Is Helping Prevent Fraud Through Faster Detection

April 10, 2018 No Comments

Featured article by Calvin Paige, Independent Technology Author

Brain 300x187 How Machine Learning Is Helping Prevent Fraud Through Faster Detection

The Federal Trade Commission fielded 1.1 million complaints of fraud in 2017, according to the latest FTC Consumer Sentinel Network Data Book. Consumers lost $905 million to fraud last year, an increase of $63 million from 2016. Over one in five people who reported a complaint experienced a financial loss, with a median loss of $429.

To fight this wave of fraud, one tool merchants and financial providers are turning to is machine learning, empowered by artificial intelligence (AI). Here’s how AI-powered machine learning is helping companies improve their efforts to fight fraud.

Current Fraud Prevention Systems and Their Limitations

While use of automated fraud review techniques is growing, most companies still rely primarily on manual review, the latest CyberSource Online Fraud Benchmark Report indicates. Address verification services, card verification numbers and device fingerprinting are rated the three most effective fraud detection tools that North American businesses currently use, the report says. However, for the most part, these tools are used in conjunction with manual reviews.

Nearly 8 in 10 businesses conduct manual fraud reviews, and companies that conduct manual reviews use them to analyze one in four orders. Nearly 9 in 10 orders that are reviewed manually end up getting approved, an indicator that companies are spending more time on manual reviews than they need to. CyberSource’s report concluded that companies could detect fraud more efficiently by adopting automated tools.

How Machine Learning Improves Fraud Detection

Machine learning represents a solution to the need for integrating automated efficiency into the fraud review process. Machine learning uses artificial intelligence to exploit the fact that fraudulent transactions exhibit different patterns of behavior than traditional ones. For instance, an identity thief may attempt to have money or goods shipped to a different address than that of the legitimate account holder.

Traditional artificial intelligence uses preprogrammed algorithms to predict the characteristics of fraudulent transactions, so that when potential fraud is detected, a review is triggered. Machine learning takes this a step further by using statistical analysis to learn fraudulent patterns from actual cases, rather than relying on preprogrammed patterns.

This allows machine learning to detect fraud far faster than manual review can — in microseconds. This is especially useful for merchants who have to deal with large volumes of transactions. It also allows companies to review more transactions with a smaller workforce, thus improving efficiency.

Machine Learning Fraud Detection Platforms

Artificial intelligence and machine learning traditionally require significant computing resources, exceeding the limits of a typical mobile device. For example, Google developed the TensorFlow library for its internal machine learning, and later made the software available to private companies for use on the Google Cloud platform, allowing businesses to use Google’s cloud resources for machine learning even if they lack sufficient local resources.

However, relying on the cloud requires communication between local and cloud resources, which slows down processing time, reducing efficiency. To overcome this limitation, smartphone component manufacturers have developed on-device machine learning to enable artificial intelligence to be run directly on mobile devices without waiting for a cloud connection. For instance, Qualcomm has developed a new artificial intelligence platform that allows mobile devices to perform on-device operations that require AI, including security operations such as biometric user authentication and hardware-based malware detection.

This type of technology is empowering apps such as First Data’s Fraud Detect, which uses machine learning to enable merchants to detect fraud before it occurs at cash registers, gas pumps, online and on mobile devices and apps. Gartner predicts that within five years, this type of AI technology will revolutionize fraud detection.

Machine learning is helping merchants overcome the limits of traditional fraud detection methods by automating the process of identifying fraudulent transaction patterns. This is enabling merchants to detect fraud faster, to process high-volume transactions more securely and to prevent fraud more efficiently. Both cloud platforms and on-device platforms now support machine learning, allowing fraud detection apps to be deployed anywhere to prevent fraudulent activity wherever it occurs.



Leave a Reply