From Input to Insight: Detecting Tone Through Machine Learning

FROM THE RESOURCE GUIDE

Jeremy R. Clopton, CFE, CPA, ACDA, CIDA
Director, Big Data & Analytics, Digital Forensics, BKD, LLP and
H. Bryan Callahan, CFE, CPA/CFF, CVA
Director, Forensics & Valuation Services, BKD, LLP

Big data. Analytics. Machine learning. Artificial intelligence. These topics, and many others, are being used with regularity in all aspects of business — from marketing and operations to recruiting and retention. They should also be topics used regularly when discussing fraud examination techniques. According to the ACFE's 2016 Report to the Nations, proactive data monitoring and analysis were associated with the highest reduction in both median loss and median duration compared to all other anti-fraud controls.

In a recent case, analytics and machine learning were applied to the analysis of a variety of textual data sources. Much of what occurred in the scheme was off-book — not recorded in the company’s financial statements. Without transactional data to rely on, our examiners leveraged other data to use in the investigation and provided information to enhance the interview process.

A large company became aware of a potential theft scheme involving the IT director and some of his direct reports. The allegations were brought to the company’s attention by a whistleblower who had previously been terminated. The individuals involved in the scheme were taking old IT equipment that still had value and selling it on eBay. The user ID used to sell the equipment was in the company’s name, though it was not in the company’s control. Rather, the IT director linked his personal PayPal account to the “company” eBay account. All payments that came through the account deposited directly to his personal account, never remitting funds to the company.

Typically, transactional-based analytics would have been the starting point. However, without transactions to analyze, examiners turned to email and instant messages of the IT department personnel. The first approach — keyword searching — did not net much in the way of direct evidence. 

The second approach — tone detection — identified a number of instant messages between the IT director and a supervisor which had a conspiratorial tone (other common tones in examinations include nervous, evasive, anxious and intimate). The topic of those communications was the eBay scheme.

In addition, tone detection also identified a couple of emails between the IT director and some female colleagues that may have been a little too “friendly” for normal professional relationships. While these were not used in this investigation, these types of results can be useful in lawsuits and investigations involving sexual harassment.

Armed with text messages, examiners interviewed the IT director who confessed to the scheme. Both machine learning — in this case specifically tone detection — and traditional analytics using keyword searching were used to successfully uncover the scheme at hand.

These topics and more are covered in the ACFE’s 2-day course, Using Data Analytics to Detect Fraud. Working through the data analysis process and assessing case studies from a data perspective, the course will help you:

  • Focus on the analytics process to successfully apply analytics in your examinations.
  • Learn the fundamental data analysis techniques and how to perform them in a variety of software solutions.
  • Learn about advanced analytics techniques, including text analytics, visual analytics and predictive modeling.
  • Strategize how to apply analytics in specific fraud schemes and develop a framework for
  • that application.

Transactions, communications, technology and other assets continue to generate more data every day. The use of analytics, machine learning, artificial intelligence and other advanced analytics methods will help anti-fraud professionals evolve their methods to keep up with the complex occupational fraud landscape.

You can read more about this course and more events and seminars in our latest Resource Guide.

Highlights from the 2017 ACFE Fraud Conference Europe

AUTHOR'S POST

Mandy Moody, CFE
ACFE Content Manager

Last week, more than 250 fraud fighters from the European region gathered in London to discuss the latest in fraud examination techniques, ethics and more. In addition to a presentation by convicted UBS trader, Kweku Adoboli (read the New York Times article, "A Rogue Trader Blames the System, but Not All Are Persuaded"), attendees walked away with actionable items to include in their own daily activities for preventing and detecting fraud. A few of the highlights include: 

When and how to approach law enforcement and prosecution
A panel of investigators discussed tips for companies and individuals to follow when reporting fraud to law enforcement:

  • Include a well-written report that defines intent. As Pennings said, “It is easier to conduct an investigation once there is a clear sign of intent. It has to hold water that it was intentional and not an accident.”
  • Make sure there is a clear collection of evidence. According to Pennings, you can’t recreate a path of evidence after the fact. It may be best to train your employees on how to forensically collect evidence at a crime scene.
  • Complete and submit a report that tells a cohesive story. Felton explained that his reports are only about half a page, so companies have to tell that story succinctly and efficiently. “If it is not clear in your head, then you have got a problem,” Felton said. “When you can’t even read through it and can’t understand it, you have got a problem. Sometimes I can’t find an offense.”

Read the full recap.

A strong ethical framework is good for business
Laura Davies, Director of Fraud at Huntswood, explained that understanding the current global landscape and putting an emphasis on culture is a step in the right direction for anti-fraud experts and organizations. Davies shared examples of recent scandals where consumer trust has been deeply shaken. Most recently, Volkswagen (VW) has been dealing with the fallout from their diesel emissions scandal. Read the full recap.

Using a risk-based pre-employment screening
Previous criminal activity can be hard to find and can put organizations at considerable risk. There are checklist recommendations that usually provide a list of information sources that should be accessed during a screening — they tell you where you can find information on an applicant. But, according to Bernhard Maier, CII, director of BM-Investigations E.U., many of these lists come from the U.S. and aren't multi-jurisdictional. Read the full recap.

Attendees also enjoyed keynote presentations by Clare Rewcastle Brown, the Editor-in-Chief of the Sarawak Report and the investigative journalist who reported on the Malaysian 1MDB corruption scandal, and Mark Livschitz, a recognized AML attorney. You can find more in-depth coverage of the conference at FraudConferenceNews.com

3 Tough Life Lessons From the Bernie Madoff Ponzi Scheme

FROM FRAUD MAGAZINE

Courtney Howell, ACFE Community Manager

On Sept. 9, 2016, Audible released the first episode of “Ponzi Supernova: Madoff Speaks,” a six-part series hosted and reported by Steve Fishman. The series focuses on the $65 billion Ponzi scheme at Bernard L. Madoff Investment Securities LLC, spearheaded by Madoff, which crumbled with the 2008 financial crisis. The last episode aired in February, and I waited until they were all available before diving in and devouring them in less than two days. As many have said before me, this is a great series for fans of the “Serial” podcast, but for anti-fraud professionals this also serves as an in-depth look into a disturbing case of widespread, unchecked fraud.

The series starts with Fishman’s exclusive telephone interviews with an imprisoned Madoff. It then works through the mechanics of the scheme and finishes with Fishman speaking to both perpetrators and victims of the crime. The most shocking aspect of the Madoff scheme is the sheer scope of it. When I first heard about it years ago, the two biggest questions I found myself asking were “how?” and “why?” But as I continued to listen, and as Fishman interviewed several peripherally involved in the scheme, I found myself asking a much tougher question — what would I do if I’d been an investor interested in Madoff’s scheme?

With that in mind, I’ve put together a list of three tough life lessons from the Madoff Ponzi scheme that fraud fighters can implement right now to improve future fraud examinations.

Lesson No. 1: Trust your instincts and don’t ignore red flags

Cynthia Keuppers, one of Fishman’s investigation subjects, once worked in the investment world and now owns a Japanese-Brazilian fast-food restaurant called Uma Temakeria. Fishman wanted to know how a scheme of this magnitude could persist for such a long period of time (more than 40 years) without anyone detecting it. This led him to Keuppers.

In 2006, Keuppers worked for Presidio, a wealth management firm in San Francisco. Some clients came to her asking about Madoff’s fund — they’d heard good things and wanted her to take a look at it. Initially she was impressed by the consistency of the returns, but before she could recommend buying in, she wanted to make sure she understood how the Madoff investment strategy worked.

“If I can’t understand it, I’m not someone that sort of says, ‘Well, if I can’t get it, somebody else must be able to get it, and I’m just not going to get there,’ ” Keuppers tells Fishman. “You have to be able to understand all the way down to where something is coming from. You might have to do a little work to get there, but you should be able to understand what a certain driver is.” This unwavering certainty in her own skills and knowledge is notable. When faced with such a highly regarded investment entity, she didn’t back down or doubt herself.

Keuppers wasn’t dealing with Madoff himself. She was working with Fairfield, the largest of the Madoff feeder funds. She had a meeting with Fairfield’s chief risk officer, and she asked him question after question, trying to drill down into the Madoff strategy. She didn’t need to understand the secret to Madoff’s success. She just needed to know why and how it was so consistent. Her objective was to see data from five years ago, but as Fishman reports, “Fairfield wouldn’t give her anything — red flag.”

The CRO assured her that he’d done the work that she was wanting to do, but he wasn’t at liberty to share the actual data with her. Basically, he wanted her to trust him. Keuppers says, “That right there was also a red flag.”

The final nail in the coffin was when Fairfield told Keuppers that the Madoff fund was closed and wasn’t accepting more investors, but he liked working with Presidio so much that they’d be willing to sell some of its holdings to Keuppers’ investors. You can hear the skepticism in Keuppers’ voice when she rhetorically asks, “Why would you sell me something that you’re not going to sell to anyone else? What makes me special?” Then she laughs and says, “Usually, you’re not that special in this industry.”

Although it wasn’t the information she was looking for, Keuppers had all the information she needed to decide. She advised against investing in the Madoff fund. As anti-fraud professionals, you might be put in similar situations. Clients push you one way, but your gut tells you to go in another. How do you make the hard decision when you know it will disappoint, or even anger, someone?

In “4 Reasons Why People Ignore Red Flags,” Jeffrey Aucoin, CFE, says that trust “is probably the biggest reason why owners and executives ignore red flags.” Keuppers made the right decision not recommending the Madoff shares to her client. Keuppers couldn’t place her trust in the validity of this investment because of Fairfield’s lack of transparency. It might have been a disappointment to her clients in the short run, but in the long run, I’m going to safely assume they’re happy with her recommendation.

Want more? Read Courtney's other two life lessons in the full article on Fraud-Magazine.com.