What Is Behavioral Analysis in Fraud Detection?

Behavioral analysis is a scientific discipline that examines patterns of behavior. In fraud detection and prevention, behavioral analysis uses patterns of behavior to identify and flag possible incidents of fraud or other criminal activity – and it is an increasingly popular way to detect fraud. It can make use of a host of behavioral data to flag anomalies and unusual transactions.

Although, theoretically speaking, it can be done manually, behavioral analysis techniques will often involve automation as well as machine learning to scale. Fraud detection via machine learning can achieve detection accuracy rates of over 90%, according to a University of Jakarta study.

Over time, customer interaction with apps and websites builds a pattern, and ultimately a profile, of normal and expected behavior. This can include a vast range of data: the times of day that individual(s) usually log on, the type of transactions they typically make, and even the nuances of how they tend to use a mouse and keyboard.

Such factors form an area where algorithms can excel. Algorithms are capable of spotting patterns that humans would miss, and behavioral analysis can spot anomalies both on an individual level and across an organization. As well as being used to detect fraudulent activity from cybercriminals, behavioral analysis is also used to detect fraud and unusual behavior by staff teams.

What Is Applied Behavioral Analysis?

Applied behavioral analysis is a sometimes controversial technique that’s also referred to as behavioral engineering. It goes beyond the analysis of behavior and seeks to modify it.

Applied behavioral analysis is used in a wide range of fields, from dementia therapy to brain injury rehabilitation and the management of substance abuse.

It has no direct relevance to behavioral analysis in fraud, at least at present.

How Does Behavioral Analysis Work in Fraud Detection?

Behavioral analysis in fraud detection begins with the collection of data via user activity monitoring. Systems log a wide range of data points that can include or expand upon:

• users’ locations and IP address details
• devices used to connect and how long for
• use of VPNs or proxies
• browser and system configuration
• typical purchase patterns
• usual transaction values
• usual cards used
• the times of day when logons typically occur

The above list is far from exhaustive, and it can also overlap with the territory of biometrics, to a certain extent. Modern behavioral analysis systems can record and analyze typing cadence, mouse use, screen display configuration, and much more.

Once a system has data to work with, it can then spot when user behavior doesn’t match up with expected patterns.

For example, let’s say that we have established that a typical online shopper will usually browse an online clothes retailer at a relatively unrushed pace – perhaps comparing prices, checking reviews, and pondering clothes sizes and colors. However, a fraudster seeking to maximize the use of a cloned card before it’s canceled would likely head straight for high-value items without much delay.

Algorithms can also be designed to improve the quality of the behavioral data that they process. The higher the quality (and often the quantity) of data that is gathered from legitimate users, the more easily systems can flag suspicious transactions.

An Example of Behavioral Analysis

In April 2021, behavioral analysis techniques prevented a fraud attack on a woman’s National Australia Bank account.

Even though the login used the correct username and password, the bank’s software solution noticed anomalies: The (ostensible) customer didn’t use the mouse in the same way as normal and didn’t cut and paste using the same techniques.

Taking the above into account, when the system detected a request to raise the woman’s transaction limit from $20,000 to $100,000, it was flagged as suspicious.

In this example, behavioral analysis was able to identify the suspicious behavior despite the credentials being correct and the customer seemingly using the same device and browser as usual.

Why Is Behavioral Analysis Important?

Behavioral analysis is a valuable weapon in the armory of automated fraud detection. The sheer volume of online transactions means that it’s not practical or feasible for companies to have humans monitoring everything in real time.

The number of digital buyers rises every year, reaching 2.14 billion in 2021. And it will continue to do so as the developing world becomes more connected. Inevitably, more connected people means more money in the system and more attempted fraud. Behavioral analysis is just one tool to combat a cybercrime industry estimated to cost businesses $10.5 trillion per year by 2025.

Meanwhile, online fraudsters are utilizing increasingly sophisticated tools to spoof devices, locations, and identities, so we cannot rely on such data alone. Behavioral analysis offers a wealth of new data points to monitor in order to catch suspicious user activity, which are more difficult to falsify as it is not readily known how exactly the average user interacts with each interface.

Due to its nature, this type of analysis is excellent at catching bot and botnet attacks, where scripts perform repetitive actions with little variation.

Types of Behavioral Analysis

Behavioral analysis solutions can incorporate several different types of analysis. These can help to combat different styles and profiles of fraud. These include:

Transaction Profiles

Transaction profiles can be formed by machine learning to build a picture of each customer’s typical account activity, based on a variety of factors. Especially in the context of ecommerce, three chief factors that form users’ transaction profiles are their typing cadence, browsing activity, and cursor movements.

Typing cadence: The rhythm and speed of people’s typing habits (such as when they enter their bank details) help to build a transaction profile that is specific to the individual. From there, we can flag whenever observed cadence deviates from this significantly.

Browsing activity: The webpage buttons that a user clicks on, the time that they tend to spend on a page, and their overall user journey (i.e. how they interact with and experience their online browsing) all form distinct transactional behavioral traits from person to person.

Cursor movements: The ways that each visitor utilizes their mouse or trackpad are very nuanced, and mouse tracking software can help the process of behavioral analysis accordingly. For instance, imagine a customer who normally uses their cursor to help them read through online text. If that customer suddenly stops using it in this way, this can be considered evidence of an imposter – enough so to trigger additional verification, perhaps.

Merchant Profiles

Working across a shop or other purchasing platform, merchant profiles can help to analyze how average customers behave on a site or app. Merchant profiles are formed of the data that vendors have on their customers, such as their choice of bank and their typical shipping address.

When someone exhibits transaction behavior that doesn’t align with the usual activity of the customers, the behavioral software will send an alert that calls for more scrutiny into the person’s actions and profile, in order to establish whether they are legitimate or not.

Behavioral analysis may also make use of and combine more fraud prevention techniques such as velocity checks and device fingerprinting as sources of data.

Regression Analysis

This technique involves taking the typical behavior of a specific demographic (e.g. online shoppers in their 20s) and considering whether a person’s activity fits in with such behavior.

If it does not correspond to the demographic they are known to belong to, they may be flagged as suspicious because regression analysis is able to detect certain anomalies. This is achieved by factoring in both dependent and independent variables and using the resultant data to determine whether an outlier applies.

How Does Behavioral Analysis Help Fight Fraud?

Behavioral analysis helps to highlight identity fraud, as well as catch criminals who have managed to convincingly imitate other signs of a legitimate customer. For example, if a criminal has obtained a victim’s authentication credentials, behavioral analysis can still raise a red flag of suspicion, because the two will act in different ways once signed in. The same goes for when their browser and device profile seem convincing enough.

Importantly, this is one type of analysis that can make use of behavioral and other data. Realistically speaking, sophisticated fraud prevention software will gather and monitor plenty of data using different sources and techniques and analyze it in different ways before flagging an individual as legitimate or suspicious.

This will typically involve using use rulesets and risk scores to ascertain when transactions should be rejected outright, or tagged for manual review. 

Given the popularity of phishing and impersonation fraud, as well as the sheer number and variety of tools and scripts criminals utilize for their nefarious goals, behavioral analysis gives organizations plenty of additional opportunities to prevent fraud and be more confident in who they do business with.