Implementation of Biometrics at ATMs

SujanParthasaradhi, Director of Biometric Applications, APAC, Biometrics, HID Global

One of the most widely used biometric modalities at the ATM is fingerprint biometrics.  It has been at use for a long time, and iseasyto use, performs well, offers reliable interoperability and the proven ability to combat fraud, and is delivered at low cost.While some may fear that fingerprint characteristics can easily be transferred, socially engineered or guessed as what has been portrayed in the movies, this is simply not true. Some of the greatest benefits of fingerprint biometrics is how universal a form of personal identity it is, eliminating any prerequisites related to literacy or language, race or gender.  A touch of the finger is all that is required for user authentication. Aadhar in India is the best example of a use case of adoption of fingerprint biometrics. Out of a population on 1.1 billion, the Aadhar system has been used 4 billion times for authentication and identification purposes. There are approximately 377 million Aadhar numbers linked to bank accounts. Going forward, the connection to bank accounts will make transactions smoother and allow bank clients to move funds just by using their fingertips.

India has topped globally in adoption of biometrics techniques, says a report. “On an average, people in India (9 per cent) are three times more likely than any other country (3 per cent) surveyed to have used ‘iris recognition’ to identify themselves,” said a well-known banks’ ‘Trust in Technology’ report. There are several key focus areas for the successful implementation of biometrics technology in a secure, convenient and trusted authentication solution. The first is biometric sensor reliability, so a solution can work reliably in a wide variety of real world conditions and environments.  To meet this need many organizations have come up with technology to acquire additional data and better images.  Unique fingerprint characteristics are extracted from the surface and subsurface of the skin for superior and reliable matching performance. Key features of some of these technologies include:

• The ability to use multiple sources and types of light, combined with polarization techniques that are used to acquire information all the way down to the finger’s capillary beds and other sub-dermal structures. This is important to generate accurate biometric data with large population and wide demographics under the conditions that are common in ATM environment such as dry, humid, ambient light.

• The inclusion of field-updatable liveness detection capabilities. This is important for combatting spoof attacks, or the use of fake fingerprints.  Attackers use these “spoofs” to gain access by impersonating a legitimate user. Through the use of advanced machine learning algorithms, the detection technology can adapt and respond to new threats and spoofs as they are identified.

• Data protection is also very critical. The ability to use industry standard encryption algorithms so the biometric data coming out of the device is encrypted and not susceptible to replay or man in middle attacks. In addition, to have the ability for mutual authentication between the device and host and provide secure memory with tamper detection and response mechanism.

According to the Minister of State for Electronics and IT in India, India witnessed more than 27,000 cyber security threat incidents in the first half of 2017.Like all identifying information, biometrics data is sensitive and must be handled properly.  System designs must be architected to protect against internal threats as well as external attacks. According to a report by EY, 33% of organizations in India do not have a SOC, compared to 44% globally. 55% do not have, or have only an informal, threat intelligence program and 44% do not have, or have only an informal, vulnerability identification capability. Data must be encrypted, but there are other alternatives to consider, including optimizing user authentication for strength and reliability, and using multi-factor and even multi-modalauthentication approaches. This helps to ensure that security is maintained even if some of the identifying data is compromised. The goal is to replace passwords and improve convenience with a seamless security approach that is non-intrusive to legitimate users.

A third consideration for optimal implementation success is the use of tamper protection, and ensuring that all connections are trusted.  The chain of trust is only as strong as the weakest link. Therefore, the devices at each point of verification must be trusted, and the biometric used to authenticate the user for each transaction must interoperate with them. An example of this device-independent, trusted physical identify verification approach can be seen with solutions which can be augmented by ensuring that the physical devices, themselves, are tamper resistant.  This preserves the integrity of all transactions.Examples of such approach includesfingerprint sensors and modules, which are implemented as follows for biometric authentication:

• The device can be encryption-enabled with various tamper resistance and detection capabilities that protect the integrity of the communication between the client and the sensor.
• The solution must preserve the chain of trust, end-to-end, in order to simplify financial transactions andprevent fraud.
• There must be a way to connect the end-point device to the other systems through a cryptographically secure channel that is also protected by hardware tamper detection and response. This is necessary in order to create trust independent of other intermediate systems and networks.
• Liveness detection is required so that the trusted biometric device can scan of a finger and know the person making the transaction is the same person that enrolled their biometric fingerprint).
• If a card, smartphone, PIN, or other authentication factor is used, each must also be confirmed by the biometric associated with the specific individual. This must be performed as part of a robust identity-proofing process at enrollment, to ensure that true identity verification has been accomplished in a trusted manner.

Biometric solutions that support this combination of tamper protection and trusted endpoint connectionshave been successfully deployed in ATMs worldwide.  This includes Brazil, where one billion annual transactions are authenticated using this approach.

Deploying biometrics solutions with tamper protection and trusted endpoint connections results in fraud reduction and other cost savings for a strong ROI. Many banks report that measurable reductions of fraud levels have met or exceeded expectations. Further, the ease of the process and attractiveness of biometric authentication has enabled banks to accelerate their enrollment initiatives, securing more transactions against fraud with biometric authentication. Finally, some banks have shown an authentication service rate for such technology sensors are higher than that of the typical card-plus-PIN authentication model.

The fourth consideration for successful implementation of biometrics authentication solutions is scenario testing.  This is required to evaluate biometric technologies across real-life environments and applications. Any well-designed system must provide a very positive user experience while maintaining required levels of security and reliability. Systems that make access difficult are frustrating for legitimate users, who then take their business elsewhere. Usability can be ascertained through scenario testing.

The fifth element in a successful implementation is the ability to link mobile credentials to a biometric identity.  It is important to remember that digital credentials are aliases for one’s true identity.  Credentials stored on personal devices must not only be authenticated, they also must be linked back to a true identity with biometrics. When credentials and digital aliases are bound to a unique individual’s true identity, a rich set of new trusted applications and services are enabled. No single security technology is 100 percent reliable; making multi-factor authentication more seamless and non-intrusive with biometrics improves and protects the user’s true identity and business transactions. And unless and until all systems are designed to support biometric authentication, both the user and vendor are exposed.

The sixth implementation element is the robustness of the biometric template.

Enhanced biometric templates, or “super templates” can be created and enforced as needed, combining biometric data with other information. This other information could be an OTP or other out-of-band data, and makes it possible for the system to recognize and reject a biometric template; for instance, a template that was created from a stolen fingerprint image. Taking this approach would enable a bank to protect user identity, and the templates could reside on a card.  It also could be placed on a chip or in a smartphone or personal wearable. Some banks already enable multi-factor authentication in this manner, mandating that a biometric and some other form of data must be provided. For even better security, a bank could enroll biometric data, and then the template could be signed and encrypted with unique or closed-system data.

Such a template could also include a combination of both standard/interoperable data as well as proprietary information. An example of this type of approach is the Secure Identity Object™ (SIO).  An SIO is a data model for storing and transporting identity information, and can be deployed in contactless and contact smart cards, smartphones, USB tokens and other form factors.  Each of these items and their data are associated exclusively with the owner’s identity, and the SIO is digitally signed using proven cryptographic techniques. Biometric data, computer log-on data, and data for other secure identity applications can all be added, encrypted, and signed, and then all content is secured with a wrapper and bound to the device with another signature.

Privacy Protection is the seventh element that must be considered for implementation success, and it starts with system design.  Ifbiometric data can be stored directly on a personal device, there is no need for the vulnerabilities of a local database or network connection. For even better privacy protection, encryption and tamper resistant devices are important for ensuring that private biometric, biographic and transactional data cannot be intercepted. Properly designed biometric solutions can also prevent attacks by fraudsters, whilekeeping control of a true identity its owner’s hands.