Yann LeCun, the French-American AI scientist has aptly quoted, “Our intelligence is what makes us human, and AI is an extension of that quality.”
Most industries on digital platforms have shown concern over fraud management. This has been a painful task across the elements of the financial services industry. The number of transactions/ taxpayers has increased in the system in the past few years. This has led to the sophistication of fraudulent activities. It is tough for regulators and federal agencies to verify identities and transactions.
But all is not lost! It is possible to apply several digital preventive measures to combat this problem. Early detection of tax fraud and suspicious activities is the key. Effective decisions can be made in two ways: identifying patterns of the transactions of a taxpayer and classifying the distinguishing characteristics of good and bad behavior of the taxpayers.
To be one step ahead in fraud detection, we need to consider the use of advanced analytics. Machine Learning (ML) and AI are good uses for this purpose. A Fraud Detection Solution can have significant benefits to financial institutions.
The Technology – Need For AI in Forensics
Banks and NBFCs have undergone digital transformation over the past few years and owing to the pandemic they are undergoing a massive digitisation. This is to cope with the demand of digital financial services since the last decade. Many major banks need extensive background checks and digital verification for their customers. AI serves as the logical choice for faster decision-making and higher efficiency.
Applications of Biometrics for Financial Industry
Biometrics have gained fame in the financial industry. The technology provides enhanced security and user experience.
The process of biometric authentication is as follows:
First, a sensor captures the user’s biometric data
It is then converted to a binary code and encrypted
The next step involves storage within a predefined database as a template
Advanced AI algorithms compare the biometric data patterns against the stored records to grant access to the user
Classification Of Biometrics Used For Authentication
Fingerprint scanning is an efficient & robust biometric authentication method. It analyzes the unique features from the user’s finger impressions. This includes ridges, curves, whorls, loops and similar identifiers of the finger. Today, fingerprint scan is a standard practice to identify a user. It is useful for carrying out a financial transaction.
In 2018, Mastercard issued biometric payment cards with an embedded sensor. It scans, processes, and verifies the identity of the cardholder.
There are several other instances of fingerprint biometrics. Secure biometric ATMs and mobile payment systems like Google Pay are some. Forgery & duplication is impossible. This reduces the probability of unauthorized access to financial assets.
Iris recognition is another efficient biometric authentication method. It has extremely low chances of false matching.
Implementation involves scanning specific areas of the eye including the iris, retina, or sclera.
For retina and sclera-based authentication, the person’s unique eye blood vessel patterns are used. These measures are not used often as they need an intricate mapping of the eye. This may cause discomfort to the user.
Iris recognition is the most convenient eye print authentication method. The user is verified even if they wear contact lenses or eyeglasses.
Banks have started to use iris authentication solutions. For example, Wells Fargo mobile banking uses a quick and easy login via eye print scanning.
According to a report By MarketWatch, the global iris recognition market is valued approximately at USD 2 billion in 2018 and is anticipated to grow with a healthy growth rate of more than 13.20% over the forecast period 2019-2026.
The Indian Government is exploring new ways of authentication for payments apps like Google Pay and PhonePe. In recent news by Business Insider, iris recognition will be a part of securing payments apps for users.
the geometric approach – this is applicable on unique identifiers for facial features
The statistical photometric approach. This assigns values to image data. Subsequently, this is compared against the templates which are kept in specific databases.
Facial recognition is another dependable biometric authentication method. It remains unaffected due to lighting or the person’s makeup. It can also identify the user from various viewing angles, including a profile view. Modern banks use this for faster customer identification during the eKYC (Know Your Customer) process.
Voice is unique for individuals and diverse for each person like the fingerprint or iris.
Voice of a person is distinguished by:
The respiration rate
specific mouth and jaw movements
A voice pattern with distinct pitch, dynamics, and intensity separates an individual voiceprint.
There are two main categories of vocal authentication:
Text-independent – This involves the analysis of any speech content provided by the user.
Text-dependent – this requires the user to utter a randomized or particular passphrase.
Vocal biometrics are highly efficient because they cannot be impersonated. The system does not confuse the user’s natural speech with a recording, and background noises do not affect its performance.
One key advantage of vocal biometrics is its user-friendliness. Vocal biometrics have no physical existence. It offers convenience as well as easy accessibility via smartphones. It can also be integrated with other devices, like smart vehicles or home appliances. Banks can also install vocal authentication to improve cost-efficiency. It can also be used to reduce the time spent for user verification – resulting in extra savings.
Newer Approach To Biometrics For Digital Forensics in Finance Sector
Passive Biometrics is a new and upcoming trend in the field of biometric verification. Passive biometrics use a concealed AI-driven security suite that consistently authenticates the user. This is done by analyzing the probable ways of device interaction, for example, smartphone or tablet keystrokes, swiping patterns, and scrolling speed.
The data is scrutinized by a multitude of parameters, which reduces unauthorized access. Banking apps which make use of passive biometrics can provide seamless authentication and frictionless experience for the user.
Banks can detect malware bots, fraudulent transactions to reduce the rates of false positives. This is due to the capability of identifying the user behavior in real-time.
A prominent use case of passive biometrics is Mastercard’s NuData Security. It utilizes four levels of security to prevent customer account takeovers and cyber-attacks.
Some systems are now designed to capture behavioural patterns of a human being. These patterns include writing rhythm, typing speed, how an individual surfs online & how we move a mouse or cursor. Researchers call it “Stylometrics”.
Postural-Information Based Biometrics
We have all heard about Motion Capture technology. It allows the movement and posture of real world people to be digitally inserted into animated movies. We have also seen how posture biometrics are recorded for a conference call across the galaxy in countless movies and PC games.
So can movement or position-based biometric data be used for digital forensics?
This new method of biometric authentication is still being researched for its potential applications. In the UK, a paper by James Ward, Chris Riley & Graham Johnson of the NCR Financial Solutions Ltd highlights the need for this technology. The paper also mentions how the placement of physical devices and their condition impacts the level of security.
Banking instruments like ATM Machines or self-help kiosks require something more robust. These instruments demand for a ‘walk-up-and-use’ system. This can accommodate a wide range of users with little or no training, supervision or external guidance.
In this process,the behavioural posture that is unique to the user is recorded for identification. This can include standing or sitting postures as well. While this is still in the experimental stage, the future looks bright for this particular segment of biometrics.
Application of Forgery Recognition in Financial Industry
Optical image recognition tools that are accessible today have become widespread in name and use. With AI-enabled systems, this can go to the next level. The system can recognize textual patterns in scanned images or in actual documents. It uses deep learning algorithms to analyze it against the original signature. This helps identify even the smallest variations.
However, there are situations where even the original signature owner might fail to reproduce a replica of his/her signature. AI technology is engineered to identify and measure the degree of mismatch. It determines whether it is, in fact, an authentic signature that has a minor difference from the original one. This greatly improves reliability as it does not cause inconvenience for users who are genuine signatories. 
Need For Image Forensics
Today, altering or defacing digital images via software is a simple ordeal at a very low cost; thus any individual can synthesize a fake picture. In the banking sector, image authentication is a strong requirement. This is crucial to your photograph as basically – “Your face is your identity”. Using AI, the exact manipulation that is made can be easily identified to the minute detail which was altered to forge the image.
Most NBFCs utilise video video-based authentication services for onboarding and customer verification. There was a lot of debate with video verification in banking. With the introduction of VideoKYC regulation by the RBI, the need for forgery detection in this category is soon to become the need of the hour.
The Future Of Digital Forensics
With a majority of business models shifting towards digital infrastructure, a fraud management system is the need of the hour. The need for forensics is not only limited to the financial industry. Digitisation has become widespread in India. Several service providers, B2B & B2C companies are looking for a digital fraud management solution. Fintechs have the opportunity to cater to this business requirement, which is bound to gain more popularity in future.
Article Contributed by Ankit Ratan is the co-founder of Signzy, a No-code AI platform for financial services.