Q&A

1. How is D4FLY different from other biometric-based identity verification security systems?

D4FLY research includes combinations of biometrics to advance the state of the art of on-the-move identity verification systems targeted at specific scenarios in border crossing environments for volunteer participants, which have given their informed consent.

D4FLY research will include new 3D facial and iris verification technology and the combination of these using new camera technology to overcome disadvantages of typical standard cameras (i.e. illumination, focus etc.). Furthermore, somatotype as a fairly new biometric is going to be researched within the D4FLY concept.

2. What does “real-on-the-move” border crossing experience means for travelers?

With “real-on-the move” we refer to the term “real-time” (i.e. fast) identity verification of travelers. The goal of the project is to be able to provide solutions that can verify the identity of a traveler as he/she passes through a specifically designed area, monitored by cameras, while movement flow remains uninterrupted.

3. How do we envision the stages for travelers using the D4FLY system during Border Control Security check, including all biometric checks that he/she would have to pass?

The first step in an envisioned future system based on the D4FLY research would be for travelers who voluntarily participate, to start the process by using their passports and biometrics to enroll at an enrollment kiosk. The next step would be, at the traveler’s destination, to walk through a dedicated corridor area with cameras set-up. The captured data would be compared with the data provided during enrolment to verify the specific identity of the traveler, while (s)he is passing the corridor. This process would enable a seamless border crossing experience for travelers, not being required to stop or queue in front of a booth.

4. What kind of datasets dos the researchers employ in the development of D4FLY biometric identity verification technology?

D4FLY, only uses databases that conform to the GDPR and EU Member States’ laws.

The D4FLY researchers initially employ publicly available databases, after officially obtaining them from their creators. These databases have been created by academic organizations with the purpose of being shared only for research purposes and as benchmarks in the research community. They come along with a License of Agreement, signed by the database creator and the party receiving and using the database.

In addition to the aforementioned, publicly available databases, a specific D4FLY database will be created within the context of the project. The database will contain data from members of the consortium, as well as other volunteers based on their informed consent according to the GDPR.

5. Facial recognition technology has been found to include racial and gender biases that disproportionately misidentify people of colour and women. How are we dealing with these two crucial issues in the development of D4FLY technologies?

These are critical issues associated with automated facial recognition technologies and ones we in the consortium take very seriously. Our ethics partner (Trilateral Research) has researched these issues in depth and continues to discuss with partners the nature of discrimination, vulnerability, and implicit bias as they relate especially to racism, sexism, and gender bias. The technology partners are using as diverse a dataset as possible for developing and training D4FLY’s facial recognition tools in order to identify and mitigate potential biases across sex and ethnicity.

6. Facial recognition technology is currently considered unreliable in cases involving individuals with similar characteristics. How do we deal with this issue in the development of D4FLY technologies?

The D4FLY research related to biometrics focuses on the verification of the identity of individual volunteers based on reference data, that has been provided by the volunteer. Furthermore, there are additional biometric modalities (somatotype, iris) to complement the verification system and “cover” the inherent weaknesses of pure standalone face recognition.

7. D4FLY aims to introduce tools for morphed faces detection. How does this technology work?

The concept of face morphing attacks refers to a synthesis of face images containing characteristics of two different individuals morphed into one image. Research is being done, under D4FLY action, seeking to develop tools capable to identify if a picture has been manipulated (using morphing techniques) or is a genuine image.

Face morphing detectors aim to identify morphed face images in personal identification documents. Morphing databases are being developed based on simple blending approaches as well as on more sophisticated approaches such as style-transfer and usage of generative neural networks. These databases of simple and sophisticated morphed face images are used to train detectors based on Deep Neural Networks (DNN), to detect morphed face images.

In case of a detected morphed face image, our methods should be able to justify and explain their decision to a human inspector. To achieve this, we use DNNs trained specifically for this purpose and Layer-wise Relevance Propagation (LRP). LRP is a method that identifies which regions in an image are relevant for the decision making process of a DNN.

8. How will iris and somatotype biometric technology will be employed in D4FLY?

Iris and somatotype biometrics are employed to complement the face verification technology under a unified fusion scheme. They are used in addition to the face recognition to increase the rates of verification and, among others, reduce potential biases across e.g. gender and ethnicities.

9. Why does D4FLY develop components for document analysis?

The authentication of travel and breeder documents is of high societal interest for the national and European security in order to prevent and detect illegal immigration, cross-border crime and fraud. The immigration services and border guards check, among others, documents from individuals from high-risk countries. Countermeasures against document and identity fraud are therefore considered of high common public interest.

10. What opportunities does blockchain technology offer for law enforcement authorities (incl. border control authorities)?

Blockchain technology enables a wide range of mechanisms such as digital signature, immutable data structure and decentralized time stamping. Moreover, operationality of the network does not depend on any single actor.

There are varying regulations between countries related to border control, therefore blockchain technology and smart contracts can be used as a general agreement between law enforcement authorities. Therefore, digital identity based on blockchain technology as a self-sovereign identity is a promising concept, which is being investigated further in the D4FLY project.

11. How will blockchain technology be applied in D4FLY project?

The concept of self-sovereign identity is studied in context of border control. Additionally, smart contract based document verification will be researched and evaluated.

12. Do the researched technologies enable surveillance of public squares or special areas, like e.g. airport halls?

The aim of the D4FLY project is specifically to facilitate a seamless and secure border crossing process for travelers voluntarily enrolling for this process. Biometric verification is only done during the border crossing and only against this previously enrolled data of the enrolled traveler. The research of these technologies is limited to this use case.

13. Will the traveler’s biometric data be stored permanently in a database?

The enrolled data must be stored for later use in the verification phase. This dataset will be encrypted and stored in a secure database. The data, which is captured during the verification process will be deleted immediately after verification. The traveler always has the option to opt out and have his/her stored data deleted or rendered unusable.

14. The project description also includes the research on means to identify known criminals, what is meant by that?

Currently, when travel documents are being checked at the border, background checks based on the passport data are already in use to identify registered suspects. The research in this project aims to investigate potential extensions of this concept by including some of the biometrics investigated in the project.

15. D4FLY aims also to enable new technologies in unobstrusive person identification. What does this mean for the traveller?

The D4FLY identity verification tools are unobtrusive in the sense that they offer travellers the option to consensually enroll in the D4FLY system in order to have their identity verified on-the-move at a Border Control Point (BCP) rather than waiting in line for their passport to be checked manually by a border guard. The traveller is well informed about the enrollment and verification process, before he or she consents to enroll. The D4FLY identity verification tools are designed to make identity verification process at BCPs more efficient for the traveller, while being compatible with new digital travel initiatives and complying with current legal requirements. The D4FLY tools and services are implemented following privacy by design principles and are not supporting surveillance capabilities.