Multi-input recognition: Combining different biometric information

• Author:
• Author name

  Quantumrun Foresight
• February 24, 2023

Looking for unique identifying characteristics beneath the skin’s surface is a clever way of identifying people. Hairstyles and eye colors can be changed or masked easily, but it’s near impossible for someone to change their vein structure, for example. Biometric authentication offers an added layer of security because it requires living humans.

Multi-input recognition context

Multimodal biometric systems are used more often than unimodal ones in practical applications because they don’t have the same vulnerabilities, such as being affected by data noise or spoofing. However, unimodal systems, which rely on a single source of information for identification (e.g., iris, face), are popular in government and civilian security applications despite being known to be unreliable and inefficient.

A more secure way of ensuring identity authentication is combining these unimodal systems to overcome their individual limitations. Additionally, multimodal systems can more effectively enroll users and provide greater accuracy and resistance to unauthorized access.

According to a 2017 study by the University of Bradford, designing and carrying out a multimodal biometric system is frequently challenging, and many issues that could hugely affect the outcome need to be considered. Examples of these challenges are the cost, accuracy, available resources of biometric traits, and fusion strategy being used. 

The most crucial issue for multimodal systems is choosing which biometric traits will be most effective and finding an efficient way to fuse them. In multimodal biometric systems, if the system operates in identification mode, then each classifier’s output can be seen as a rank of enrolled candidates, a list representing all possible matches sorted by the confidence level.

Disruptive impact

Multi-input recognition has been gaining popularity because of the different tools available to measure alternative biometrics. As these technologies advance, it will be possible to make identification more secure, as veins and iris patterns cannot be hacked or stolen. Several companies and research institutions are already developing multi-input tools for large-scale deployment. 

An example is the National Taiwan University of Science and Technology’s two-factor authentication system that looks at skeleton topologies and finger vein patterns. Finger vein biometrics (vascular biometrics or vein scanning) uses unique vein patterns in a person’s fingers to identify them. This method is possible because blood contains hemoglobin, which shows different colors when exposed to near-infrared or visible light. As a result, the biometric reader can scan and digitize the user’s distinct vein patterns before storing them on a secure server.

Meanwhile, Imageware, based in San Francisco, uses multiple biometrics for authentication purposes. Admins can select one biometric or a combination of biometrics when implementing the platform security measure. The types of biometrics that can be used with this service include iris recognition, facial scanning, voice identification, palm vein scanners, and fingerprint readers.

With ImageWare Systems’ multimodal biometrics, users can authenticate their identity anywhere and under any conditions. Federated login means that users don’t have to create new credentials for each business or platform because their identity is created once and moves with them. Additionally, single identities that are cross-compatible with different platforms allow for less exposure to data hacks.

Implications of multi-input recognition

Wider implications of multi-input recognition may include: 

• Population-scale improvements to cybersecurity standards as (long term) most citizens will utilize some form of multi-input recognition as a replacement to traditional passwords and physical/digital keys to secure their personal data across multiple services.
• Building security and sensitive public and private data experiencing incremental security improvements as (long-term) employees with access to sensitive locations and data will be mandated to utilize multi-input recognition systems.
• Companies deploying multi-input recognition systems that use deep neural networks (DNNs) to correctly rank and identify this different biometric information.
• Startups focusing on developing more multimodal recognition systems with various combinations, including voice-, heart-, and faceprints.
• Increased investments in securing these biometric libraries to ensure that they don’t get hacked or spoofed.
• Potential incidents of biometric information of government agencies being hacked for fraud and identity theft.
• Civic groups demanding companies to be transparent on how much biometric information they gather, how they store it, and when they use it.

Questions to comment on

• If you have tried a multimodal biometric recognition system, how easy and accurate is it?
• What are the other potential benefits of multi-input recognition systems?