January 30, 2018

Biometrics-enabled age verification systems: legal and technological considerations


In this blog, I explore various options for age verification for legal purposes, e.g., only people above 18 years old can buy alcohol. Here, I consider the legal aspects and in light of these, what technological solutions are readily available.

Principle of proportionality with regards to the use of biometrics

Principle of proportionality sets out a guideline to limit the use of evaluating, processing, and storing biometrics unless there is a clear necessity. By necessity, it should be understood as (1) when the security outweighs a possible interference with human rights and (2) other less intrusive techniques do not suffice.

Refer: The principle of proportionality applied to biometrics in France: Review of ten years of CNIL’s deliberations

The principle of proportionality was originated from the need to protect sensitive data; and this applies in certain member EU states where biometrics data is considered sensitive data.

Refer: EU Data Protection Legislation and Case-Law with Regard to Biometric Applications

Brexit may or may not change this view. Rather than being speculative, the status quo is generally preferred.

##Technological considerations

Let us pick the scenario of the retail where only adults above 18 can purchase alcohol. This is verification of credential, i.e., the right to do something. Let’s consider various solutions:

1. Document identity verification.

The vendor can verify the identity of the person via a national identity document which contains the date of birth of the person, from which the age can be derived. Countries without a unified national identity scheme (e.g, national identity card) will need alternative documents such as driving licence or other documents which can be easily forged. National identity documents often have some form tamper-proofed security to avoid forgery such as holograms, smart cards containing digital credentials, etc.

2. Biometrics – as in person identity verification.

Biometrics can certainly solve this problem; but the identity must be associated with the demographic information, stored digitally. Simply verifying John is John is not sufficient; we must know the age of John.

This implies putting in place a system, or else piggybacking on an existing national system where biometrics data has already been adopted and use. Unfortunately, this is not the case in the UK. Other developing nations such as India, Malaysia have already adopted biometrics-enabled documents which allows different commercial sectors such as banking to use them as identity verification, as a standardized means of proof of identity. However, age verification is trivial; the vendor can check the national identity document and check against the facial photo.

3. Biometrics as age estimator.

With deep learning, and other methods, it is now possible to estimate the age of a person from just facial image. There are plenty of literature here, and the accuracy of the system has reached a point that it is sufficient to give an approximate perceived age for the purpose of analysing advertising campaign. For example, it is often necessary to find out who has viewed a campaign by breaking down into age, gender, and ethnicity. With appropriate labelled training data, deep learning neural network models can now be easily created, hence delivering a classifier system that is optimised for a specific demographic segment.

Summary and open questions

In light of the above, in countries where national identity documents are not well established, biometrics as an age estimator is more likely to be adopted from the technological perspective. The technology for age verification is readily available and can be improved upon to create USPs that are attractive to certain retailers. There are, however, open-questions: Are they willing to pay for this solution? Does the USP outweigh the cost? How to quantify this ‘cost savings’? These are, in my opinion, some of the questions worth exploring further.

Cite this blog post


    @misc{ poh_2018_01_30_biometrics-age-verification,
      author = {Norman Poh},
      title = { Biometrics-enabled age verification systems: legal and technological considerations },
      howpublished = {\url{ http://normanpoh.github.com/blog/2018/01/30/biometrics-age-verification.html},
      note = "Accessed: ___TODAY___"