The Knowledge Exchange / Risk Management / Are organized fraudsters winning the big data battle?

Are organized fraudsters winning the big data battle?

A layered analytics approach complemented by high-performance computing makes it easy to find a needle in the haystack

Ellen Joyner RobinsonOrganized fraudsters love the big data battle because – for many organizations - siloed data and too much data make finding fraud as difficult as finding the same needle in a new haystack.

This is no little problem. According to a 2011 study by Norton, cybercrime costs the US an estimated $400bn annually – much more than the combined global market for illegal drugs.

Many financial institutions are challenged when looking for suspicious activity across silos, and by the inability to share data across product lines and channels . Then you compound those issues with the need to make sense of big data – especially when it comes to a split-second decision as to whether or not to stop a transaction. 

These linchpin decisions keep fraud and security executives up at night and help shape the decisions that ensure that the right systems are in place to minimize money  and reputational loss, and just as important, prevent customer attrition due to a bad decision.

Fraudsters know this, and especially organized fraudsters that can quickly move from one area of the business to another at the first sign of potential discovery. According to the 2011 Data Breach Investigations Report by Verizon, “Thirty-three percent of the time, it only takes minutes from the time that the bad guy gets a ‘point of entry’ before data ‘compromise.’” In the same report, organizations were alerted to a data breach in only 46 percent of the cases. That leaves a huge opportunity for the fraudsters.

 Some other startling statistics from the Ponemon Institute –  61 percent of financial institutions believe that only one successful fraud involving online bank accounts could destroy customer trust, and 85 percent of the customers interviewed say they would transfer their business to another bank if that happened.

That leads us back to the original question about big data: What can financial institutions do about meeting the challenge head on? Firms need the right tools in place to sort the massive amounts and kinds of data. And, organizations need high-performance analytics to quickly find the criminal or network of criminals. These analytics will provide the confidence that the decisions made from the results are the right ones. 

This can only happen if you are using the behavioral and predictive analytics where user, account and channels are being monitored to identify anomalous behavior across channels and products.  In addition, alerts must be correlated for each entity. 

For one large, global financial institution that manages more than 36 card portfolios in over 80 countries, a collaborative environment for enterprise fraud management is critical. Handling more than 3,000 business rules, advanced neural networks and unlimited data sources are a requirement for successful monitoring of suspicious activity across product and channels. Now more than 35 percent of fraud is being detected through a layered analytics strategy.

Your organization can bring big data and organized fraudsters down to size with the right tools and methodology: a layered analytics approach combined with high-performance computing.  The Aite Group said it best, “The institution with complementary layered technologies is akin to the house with a high fence, a big guard dog in the yard, and a burglar alarm inside. This provides multiple opportunities to catch the bad guys in the act, and encourages the criminals to go in search of easier prey.”

Read more about what high-performance analytics can do.

Tags: , ,
  • Facebook
  • Twitter
  • Digg
  • LinkedIn
  • email

One Comment

  1. Peter Dorrington, Director of Marketing Strategy (EMEA), SAS
    Posted May 30, 2012 at 11:47 am | Permalink

    Hello Ellen,
    Great post!
    I think that the opportunities that high-performance analytics when applied to big data and in countering fraud are tremendous:
    - 100% sampling – most fraudulent transactions are a tiny percentage of the overall number. Sampling means that, even if we are lucky enough to find one of those transactions, the nature of sampling is such that it may be some time before we come across another transaction we can link to it and begin to form a pattern. 100% sampling means that we ‘see’ everything and are therefore much more likely to spot an emerging pattern earlier.
    - multiple analytics techniques; with high-performance analytics you can use multiple detection methods on 100% of the data in the time it used to take to run just one model on a sample of the data – you are going to find patterns of behaviour your were likely to miss before
    - External data – if we take the insurance industry as an example, the number of interactions they have with their clients in any given year is quite low. However, if we include external data (for example, public domain social media data) and blend this with the internal data already have we can form a more complete picture of all clients, including fraudsters
    - Discrimination – a big issue for everyone involved in countering fraud is how to strike the balance between access and protection; we want to prevent fraud but without negatively impacting the customer experience; more models, using more data in less time inevitably leads to systems that are much better at telling the difference between a genuine customer and a fraudster and therefore more efficient at processing both.
    Finally, the same data that we use to detect fraud can be the same data that we use to manage risk, enable great marketing, manage the business and make our statutory reports; big data is an opportunity to leverage a high-performance analytics across the business, nit just departmentally.

Post a Comment

Your email is never published nor shared. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>