BPM Analytics
Tackling subscription fraud with fuzzy logic
Identity theft is not a joke. Nor is combatting subscription fraud. This is where fuzzy logic steps in.
The most prevalent fraud against communication service providers is subscription fraud, which is typically defined as a fraud where perpetrators use synthetic identities to obtain services, with no intention to pay. Advancements in technology have further increased fraudulent activities in the telecom sector. Fortunately, newer technologies in analytics, machine learning, and artificial intelligence have come to the rescue in combatting telecom fraud. Moreover, advanced analytics, like fuzzy matching, have hastened the evolution of fraud detection and prevention techniques.
What subscription fraud looks like
The 24x7 online lifestyle has led to two things – an abundance of data and an absence of privacy. Consider your mobile number, the primary source for verification of one-time-passwords (OTPs) for a host of activities such as banking or e-commerce shopping. Simply targeting mobile numbers, especially in the absence of robust KYC (know our customer) controls, can aid fraudsters in committing various financial frauds and scams, such as International Revenue Share Fraud (IRSF), SIM Box Fraud, or Interconnect Bypass Fraud.
As per a 2020 report by TM Forum, five employees of one of the biggest telecom service providers in US and two external scammers in 2019 stole multiple identities and created unauthorized accounts to purchase many mobile devices and accessories. Similar frauds have been recorded across the globe as well, where fraudsters used lost and manipulated national identities to activate postpaid mobile subscriptions, leaving various legitimate customers with hefty bills.
Common measures to control such instances have been implemented by various service providers worldwide, that have created awareness among customers about such frauds and subsequently helped increase reporting of such data breach incidents. This could mean immediately reporting IDs, if lost, biometric fingerprint verification, and so on. These techniques could, however, have certain limitations, such as confirming the identity of real customers whose biometrics are associated with a fake or stolen ID.
Such limitations have made one thing apparent. There is a requirement for operators to design a future-ready solution with improved analytical capabilities, not just to detect, but to predict fraud. An advanced technique that can establish the intention to pay or the lack of it, based on usage or demographic information. This can be achieved by processing a variety of data sources to identify the relationship between subscriptions and fraudsters through fuzzy matching or fuzzy logic.
Combatting subscription fraud
According to the survey conducted by Communications Fraud Control Association (CFCA) in 2021, 30% of service providers rely on manual KYC processes to detect and prevent subscription fraud. With only 15% of the providers using advanced automatic decisioning, there is a huge potential for process automation and analytics for real-time verifications. Thankfully, the rapid adoption of 5G use cases and connected devices, has made it easier for service providers to deploy advanced means such as fuzzy logic to nip the fraudsters in the bud.
When available data does not give clarity, analysts look for patterns or trends in the datasets. Fuzzy logic helps assist with this by spotting similarities in data. This method can recognize two pieces of texts that are approximately similar, but not the same. The fuzzy matching technique attempts to classify data from large datasets through subjective reasoning. Based on its findings it provides match scores ranging in degrees between 0 and 1, also called the degrees of possibility or truth value. The answer here is not definite a yes or a no, but different intensities or degrees of the truth. This technique is more granular and efficient compared to the deterministic approach or logistic regression which provides outputs in binary.
In technical terms, fuzzy logic has a collection of membership functions, i.e., universal truths and fuzzy rules defined in conditional statements. If the fuzzy system is correct, all the rules are corrected partially to the defined range. If the antecedent is true to some degree of membership function, then the outcome is also true to some degree. Fuzzy matching plays a crucial role in identifying recurring subscription frauds, where fraudsters might tweak small identity details such as address, name, occupation, salary details, etc., and match them against previous subscriptions to detect fraudsters early on.
Fuzzy logic helps in categorizing the data more accurately. Even if there is a mismatch or ambiguity in data, possibilities can be identified based on the degree of possibility score. Along with this, the fuzzy matching technique is more flexible, easy to build, and configurable than its analytical counterparts.
Fuzzy logic may very well be the solution that telecom companies were waiting for.