Anti-fraud – this is how to identify frauds in 2022

For another year, the most common frauds in the finance sector are fraudulent loan and credit products, and as many as 80% of leasing companies perceive an increase in the risk of their occurrence compared to the previous year.

As recently as 2020, these latter institutions were in 2nd place in terms of increased fraud risk – then banks led the way (58%). The issue of actions taken to identify fraud in the financial sector is explored in this year’s (2021) report by EY and ZPF, so in this article, we will focus on ready-made solutions. But first…

Many institutions, one challenge

The 2021 edition of the survey involved 3 main groups of respondents:

  • banks,
  • lending institutions,
  • leasing companies.

In smaller numbers, respondents were also representatives of such groups as:

  • insurance companies,
  • financial intermediary sector,
  • FinTech companies.

The report shows, among other things, that despite the slow stabilization of the level of fraud risk (answer ‘Intensity of fraud in the last 12 months – no change’ 41% in 2020 and 8 percentage points more in 2021), the threat is still high. The EY and ZPF publication indicates the aforementioned groups of financial institutions are exposed to increasingly sophisticated ways of defrauding (e.g., new fraud schemes). The challenge in efficiently detecting fraud attempts is distinguishing a customer with lower creditworthiness from a fraudster, and the AlgoLine technology suite developed by Algolytics has been helping institutions improve this process for years. Knowing these facts, let us look at the question of cost effectiveness.

Anti-fraud solutions as an investment in a stability

The report indicates that almost 40% of financial institutions recorded losses of over 500 thousand PLN (about 110 000 EUR) in 2021 alone! Even 80% of leasing companies reported a loss of more than half a million PLN. The same problem is also faced by 20% of lending institutions and 16% of banks. What is worrying is that 13% of smaller businesses do not know losses incurred due to fraud…

As you can see, organizations need an effective way to combat the fraud problem. The standard solution to the fraud problem is to learn from the already detected frauds and their attempts, and then gradually tighten the rules of granting e.g., loans. However, the data from EY and ZPF report shows that only 22% of institutions perceive their preparation for counteracting fraud as “good”. What about the remaining group of respondents?

Most respondents are “rather prepared” – which may indicate process deficiencies. What is more, it is not the issue of allocating resources that keeps Risk Management up at night. The report states that:

“33% of organizations spend more than PLN 100,000 on fighting fraud, while 8% of the surveyed institutions spend more than 500 thousand PLN.”

Interestingly, none of the respondents stated that these expenditures are too high, and in response to the question about assessing the number of funds currently allocated to anti-fraud solutions, ¼ of the respondents are willing to increase spending on anti-fraud. And no wonder – if institutions invest PLN 100,000 (about 25 000 EUR) a year in the fight against fraud, the investment can pay for itself 5 times over.

Fraud at the target of financial institutions

This fact sheds light on another statistic – 90% of leasing companies, 80% of lending institutions, and 66% of surveyed banks notice increased restrictions in the process of granting financing (in comparison to the previous year). According to Algolytics experts, it is the combination of classical scoring rules with ML algorithms and the possible intervention of a financial institution representative that is the most effective way to prevent fraud. In banks or telecoms, this form is nothing new – and yet it turns out to be necessary (especially in the perspective of newer and newer fraud methods) to constantly update the anti-fraud process. The answer to this challenge is, for example, the AlgoLine package, which enables operation in the Continuous Training Machine Learning model – where each “caught” case of fraud, or its attempt improves the fraud detection process.

A risk is inherent: using stolen data, data manipulation, fictitious data…

In the age of the open Internet and entire forums devoted to ever more clever ways to defraud, the opportunity for a stable future for the financial sector lies in intelligent anti-fraud solutions. Technology can detect a phishing attempt as early as the contact form stage. Let us take 3 example use cases into account:

  • Person 1 manipulates his data when filling out a loan application. One time the type of contract differs, then the amount of earnings…. and so on, until he gets a positive decision.
  • Person 2 completes the form with stolen data. Here, for example, he may use an illegally obtained list of personal information.
  • And person 3 tries to take out a loan using an invented identity

What countermeasures should be taken?

In case of data manipulation, it is possible to detect it by automatically triggering cross checking rules. By cross-referencing entered data with multiple sources of information each time, it becomes easier to spot attempted manipulation. Comparison with historical data proves to be invaluable. So, if a form is completed several times, e.g.:

  • from the same location,
  • from a device with the same IP,
  • and the changes introduced are different – e.g., the amount of earnings is increased,

these elements are efficiently captured by the system. Example?

Among recent cases of fraud attempts, Algolytics technology detected e.g., a married couple, who in about a year submitted 34 credit applications.

  • the couple’s source of income was changing – from pensions and benefits to employment contracts in various industries;
  • the couple’s home address was changing;
  • or marital status for both spouses – married, widowed, married
  • the total value applied for was PLN 140,000 with a total income of about PLN 5,000.

And this is just a single case!

Well, let us raise the bar – the next challenge is to detect the use of stolen data. According to Algolytics research, the verification process here must cover multiple levels, as traditional anti-fraud methods are insufficient.

  1. Firstly, the process of verification in external databases and verification of connections is enriched with the biometric aspect. This one is indispensable when someone, for example, takes the ID of an elderly family member. To put it simply: when entering the data of a 70 year old person, a 20-year-old fraudster will use the computer completely differently and at a different pace. In this case, among others, the way the device is used is one of the biometric indicators lowering the scoring, and appropriate anti-fraud rules precisely catch abnormalities deviating from the norm.
  2. Secondly, bot detection. Writing an application that completes a form with data, e.g., from a stolen list, is not a major challenge nowadays. Detection of such a bot is possible among others by verifying the speed of data entry. If, for example, each letter appears in the form at identical time intervals – this is a sign that it is time to prevent the fraud attempt.
  3. Thirdly, to further tighten the anti-fraud process, we created Know Your Customer. This SaaS service generates a set of questions about a customer’s location relative to current geographic data. Providing an incorrect (or no) answer will result in a denial of funding.

But moving on! Use Case number 3, the detection of a made-up identity, will rely on similar mechanisms as the detection of the use of stolen information. Here again, in addition to verifying data in external databases, it is also worth automating the detection of anomalies and links. In case of fraud attempts with a higher risk indicator, e.g., through the Call Center channel, it is worth to seal the process with the already mentioned Know Your Customer protection. At one point we all heard a standard question about a mother’s maiden name. Now, however we enter a higher level of fraud detection – with the use of geolocalization.

Guarding security and convenience

Just take a look at the list of 10 example questions, which will be answered quickly and correctly only by a real person living at the address provided. Questions are generated automatically – based on current demographic and geographical data. This solution will work as an element of sealing anti-fraud processes in every financial organization – from small lending institutions to banks. However, we emphasize here that these sample questions are part of the entire verification process.

  1. What municipality (name) do you live in?
  2. Is your municipality of residence an urban, rural, or town-rural municipality?
  3. What county (name) do you live in?
  4. What is the name of the largest city in the province where you live?
  5. Do you live in a single-family or multi-family building?
  6. How many apartments are there in the building you live in? Order of size e.g., up to 5, up to 20, over 50.
  7. Is there a highway within 1 km of your home?
  8. Is there a train station within 1 km of your place of residence?
  9. Is there a river within 1 km of your place of residence?
  10. Is there a cemetery within 1 km of your location?

(Psst! Check Know Your Customer here. Create the account and get your access to the free trial).

Using Algolytics technology, this process takes place in real time and is unnoticeable to the user of your website. Effective anti-fraud is based on the combination of mechanisms in such a way as to provide the convenience of use for each party – both the customer and the financial institution. When a person initiates the loan process, Algolytics technology for Risk Management collects such data as

  • Type of device used, operating system,
  • Behavioral information and user behavioral data,
  • Browser type,
  • Geolocation data,
  • Internet Service Provider information.

In the background, verification, analysis, and comparison of the incoming data with databases and historical values are performed simultaneously. Based on the result of the verification process, 3 further action scenarios are possible:

  1. Acceptance of the request;
  2. Recommending additional verification of the user;
  3. Or Rejection of the request;

What is important, the results are distributed multichannel – so that both the stationary banking point and the Risk Management Department have immediate information about a suspicious customer. And while we are on the subject of customers…

Scoring is not the limit!

Today, scoring algorithms are available in few clicks, but their results are… not necessarily satisfactory. That is why while creating technologies for financial institutions Algolytics experts significantly raised a bar for standard solutions. Considered were such aspects as:

  • All data in 1 place. The more information, the higher precision in fraud detection – and access to dozens of sources in 1 place and via 1 API is more than just time-saving convenience. The solution is easy to implement and unnoticeable for an external customer because the process takes place “in the background”. To maximize the use of this technology, Algolytics is also responsible for regular (e.g., monthly) updates of all data.
  • New fraud patterns learning. With current and historical observations, it is easier to learn algorithms for new phishing patterns – especially when a person’s behavior is compared with tens of thousands of input parameters (including behavioral or biometric data). User behavioral biometrics is done in real time – for any device, for any customer…
  • More application approvals, more sales. Other technologies only serve to reject applications – and customers who are denied move on to competitors. In addition, 1 in 6 of these people share the unpleasant experience further, e.g., on the Internet. In the case of AlgoLine, precise anti-fraud rules effectively prevent frauds and at the same time mean a higher application acceptance rate. This is because the algorithms indicate customers who repay their liabilities despite their “non-ideal” scoring. Moreover, in this case, the increase in the sales value does not require any additional work on the part of the financial institution – after the implementation of AlgoLine, the process simply starts.
  • Just one article is not enough to describe all the possible examples. 😉

    Considering the statistics from the report, the fact is that countering financial fraud is a process that requires a data-driven expert approach, and the quality of the technology significantly affects the results (and profits). So, if you want our experts to help you minimize the number of frauds in the financial industry, present demos of our proprietary solutions, and advise you on how to effectively optimize the processes already taking place – feel free to contact us.

Pin It on Pinterest