Captured source text (L5)
The robustness criterion implementing age assurance?
Where relevant, ensure the technology has been tested in a range of conditions
4.27We expect service providers to implement age assurance processes that have undergone
testing to ensure the process is highly effective in a range of conditions; e.g. poor lighting,
blurring, brightness, contrast, or positioning of the user in the image.
4.28Should service providers choose an age assurance method dependent on visual or audio
input, they should ensure that the technology underpinning that method has been tested in
multiple environments during its development, to minimise any discrepancies in the
performance of the method in actual deployment contexts.
Identify and take appropriate steps to mitigate against methods of
circumvention that are easily accessible to children and where it is reasonable
to assume that children may use them
4.29We expect service providers to identify and take appropriate steps to mitigate against
methods of circumvention that are easily accessible to children, and where it is reasonable
to assume that children may use them.
4.30Where service providers implement age assurance processes that rely on details obtained
via a user’s identification documents, mobile phone number, email address, or credit card,
we expect providers to have means of checking that the details supplied belong to the user
attempting to access the service.
4.31Verifying the ownership of a user’s details is one method of reducing the risk of a child user
circumventing the age assurance process. For example, in the case of a mobile phone
number, email address or credit card checks, this could be done through the use of a One
Time Passcode (OTP) or multi-factor authentication.
4.32Requiring a photo of the user at the point of ID upload when photo-ID matching helps to
verify that the photo ID belongs to that user. Liveness detection provides further
confidence that a child user has not uploaded a photo of an adult by ensuring that the user
undergoing the age assurance process is present at the time the check is carried out.22
4.33Liveness detection can also help to ensure that children are not using still images of adults
to pass through facial age estimation.
4.34It is possible to obtain fake forms of identification of varying degrees of sophistication. A
robust photo-ID check should not be capable of being easily circumvented, and as such, a
photo-ID method should be able to detect falsified documentation or manipulation that a
child could create or obtain. Government-issued guidance on how to prove and verify
22 Liveness detection is used to ensure that the face being analysed is not a photograph, video, or any other
form of spoofed representation. The primary goal is to prevent attackers from using static images (print attack)
or pre-recorded videos (replay attack) to trick the system into making inaccurate age estimates.
15
someone’s identity (“GPG45”) provides some useful indicators on how a document can be
scored to detect certain levels of faked documentation.23
4.35Repeating an age check could also help to increase the robustness of an age assurance
method, to prevent child access via device or account sharing and instances where children
may be mistakenly classified as adults during the initial age check.
4.36Service providers should determine whether repeated age checks are needed to secure the
robustness of their solution based on the features of their service, and if so, how often it is
appropriate to repeat an age check. For example, service providers may decide to age check
each unique visitor to a service which does not require users to create an account. When
deciding on the frequency of age checks, service providers should be mindful that data
protection law requires them to assess the necessity and proportionality of the personal
data processing and take a data protection by design approach to implementing the data
protection principles.24
Full captured verbatim text
Where relevant, ensure the technology has been tested in a range of conditions 4.27 We expect service providers to implement age assurance processes that have undergone testing to ensure the process is highly effective in a range of conditions; e.g. poor lighting, blurring, brightness, contrast, or positioning of the user in the image. 4.28 Should service providers choose an age assurance method dependent on visual or audio input, they should ensure that the technology underpinning that method has been tested in multiple environments during its development, to minimise any discrepancies in the performance of the method in actual deployment contexts. Identify and take appropriate steps to mitigate against methods of circumvention that are easily accessible to children and where it is reasonable to assume that children may use them 4.29 We expect service providers to identify and take appropriate steps to mitigate against methods of circumvention that are easily accessible to children, and where it is reasonable to assume that children may use them. 4.30 Where service providers implement age assurance processes that rely on details obtained via a user’s identification documents, mobile phone number, email address, or credit card, we expect providers to have means of checking that the details supplied belong to the user attempting to access the service. 4.31 Verifying the ownership of a user’s details is one method of reducing the risk of a child user circumventing the age assurance process. For example, in the case of a mobile phone number, email address or credit card checks, this could be done through the use of a One Time Passcode (OTP) or multi-factor authentication. 4.32 Requiring a photo of the user at the point of ID upload when photo-ID matching helps to verify that the photo ID belongs to that user. Liveness detection provides further confidence that a child user has not uploaded a photo of an adult by ensuring that the user undergoing the age assurance process is present at the time the check is carried out.22 4.33 Liveness detection can also help to ensure that children are not using still images of adults to pass through facial age estimation. 4.34 It is possible to obtain fake forms of identification of varying degrees of sophistication. A robust photo-ID check should not be capable of being easily circumvented, and as such, a photo-ID method should be able to detect falsified documentation or manipulation that a child could create or obtain. Government-issued guidance on how to prove and verify 22 Liveness detection is used to ensure that the face being analysed is not a photograph, video, or any other form of spoofed representation. The primary goal is to prevent attackers from using static images (print attack) or pre-recorded videos (replay attack) to trick the system into making inaccurate age estimates. 15 someone’s identity (“GPG45”) provides some useful indicators on how a document can be scored to detect certain levels of faked documentation.23 4.35 Repeating an age check could also help to increase the robustness of an age assurance method, to prevent child access via device or account sharing and instances where children may be mistakenly classified as adults during the initial age check. 4.36 Service providers should determine whether repeated age checks are needed to secure the robustness of their solution based on the features of their service, and if so, how often it is appropriate to repeat an age check. For example, service providers may decide to age check each unique visitor to a service which does not require users to create an account. When deciding on the frequency of age checks, service providers should be mindful that data protection law requires them to assess the necessity and proportionality of the personal data processing and take a data protection by design approach to implementing the data protection principles.24