Just start typing...

Case Study: ML-based module to detect users’ fraudulent activity

Published March 6, 2018

WaveAccess developed an anti-fraud system to protect telecom companies from fraud attacks. The solution is based on Machine Learning and fuzzy logic algorithms. The security module recognizes suspicious actions as threats.

Telecommunication companies are constantly exposed to fraud attacks. There are attempts to gain access by using fake IP addresses, stolen credit card numbers, password matching and other attacks — it is impossible to guess what the fraudsters will come up with. The suspicious actions are easily confused with the actions of a simple user, e.g., with an input error or access from a new device. Therefore, the standard software protection tools become less effective and just downloading antivirus software doesn’t work — special tools are required to study user behavior and compare it with potential threats.

WaveAccess developed a machine learning based module that helps safely protect telecom companies — the system detects attacks in time, even if more than a million accounts are logging in. WaveAccess experts analyzed the most common threats, defining and formalizing 16 main threats types. For each of them an algorithm was developed that determines how a threat will be defined and dealt with. For some threats, a user’s feedback is requested, which determines the reaction to a threat.

WaveAccess_anti-fraud system_the model of reactions to threats

The model of reactions to threats

To keep setup and monitoring processes simple, WaveAccess developed the submodule with UI. It monitors event logs and security logs and sets up the Anomaly Monitor’s subsystems.

WaveAccess_anti-fraud system_security log

WaveAccess implemented a number of load tests and integration tests to ensure that the protection module can easily handle a huge amount of input data. For Integration testing, the testing utility — scripts imitator — was developed. It simulates the load on the subsystem and helps to implement integral testing of the main module using templates. This approach saved time and resources that are normally spent on manual testing. It also allowed to add the testing into Continuous Integration.

WaveAccess creates similar machine learning based modules for “smart” recommendations in online stores; the same algorithms are used by the company's clients to calculate the probability of an insured event; to predict the probability of buying and other business tasks. These solutions can also be applied for protection by examining and detecting abnormal user behavior.

Find out more about the project in our case.

Download the case

Related Services

How we process your personal data

When you submit the completed form, your personal data will be processed by WaveAccess USA. Due to our international presence, your data may be transferred and processed outside the country where you reside or are located. You have the right to withdraw your consent at any time.
Please read our Privacy Policy for more information.