A study published May 7 documented how researchers developed a method of recreating three-dimensional facial models of individuals based off their DNA.
“Facial morphology is a distinctive biometric marker, offering invaluable insights into personal identity, especially in forensic science. In the context of high-throughput sequencing, the reconstruction of 3D human facial images from DNA is becoming a revolutionary approach for identifying individuals based on unknown biological specimens,” the study said in the ‘Abstract’ section. “Inspired by artificial intelligence techniques in text-to-image synthesis, it proposes Difface, a multi-modality model designed to reconstruct 3D facial images only from DNA.”
The method can even be used as a predictive characteristic tool as well.
“Amazingly, Difface could generate 3D facial images of individuals solely from their DNA data, projecting their appearance at various future ages,” Luonan Chen, an author of the paper and a researcher with the Chinese Academy of Sciences said.
While the technique has been pioneered using only Han Chinese people, a very homogenous breed of human, expanding the technology to other human breeds is necessary to hone the technique, according to the researchers.
“Future studies should focus on this crucial extension by testing and adapting the model to encompass more genetically diverse populations. Validating Difface with datasets from multiple ethnic groups and exploring whether additional genetic loci are necessary for certain facial features will be key steps to ensuring that the model generalizes effectively across diverse populations. This broader application will not only enhance the model’s robustness but also significantly increase its utility in real-world forensic investigations and personalized medicine on a global scale,” the study said in the ‘Discussion’ section.
The technique is the latest work in the field of DNA phenotyping (identifying physical characteristics of an individual based off their genetics.
“Forensic DNA phenotyping (FDP) has emerged as a groundbreaking tool in criminal investigations, offering the potential to predict an individual’s physical appearance based solely on their genetic material. This technology aims to generate leads in cases where traditional methods have stalled, providing law enforcement with new avenues to pursue. However, the application of FDP is fraught with scientific, ethical, and legal challenges that warrant careful consideration,” Biometric Update said in February. “Parabon NanoLabs, a biotechnology company based in Reston, Virginia, has been at the forefront of FDP through its Snapshot DNA Phenotyping Service. Launched in 2015, Snapshot claims to predict various physical traits, including eye color, hair color, skin tone, and even facial morphology, from DNA samples collected at crime scenes. The company emerged from computational and bioinformatics expertise initially aimed at therapeutic applications before pivoting towards forensic applications. It has worked closely with the U.S. Department of Defense and law enforcement agencies to refine and deploy this technology.”
DNA phenotyping is not without its pitfalls however. While predicting hair or eye color is fairly straightforward, predicting facial features is much more complicated and less accurate, at least until the recent development of Difface.
“Beyond scientific validity, FDP poses substantial ethical and legal dilemmas. One major concern is the potential infringement on individual privacy rights. DNA contains sensitive information not only about an individual’s physical traits but also about their health predispositions and familial relationships. The use of genetic data to predict appearance without consent can be seen as a violation of privacy, especially when such data is obtained from public genealogy databases. The case of the Golden State Killer’s apprehension in 2018, which involved uploading crime scene DNA to a public genealogy database, sparked a debate over the balance between public safety and individual privacy rights,” Biometric Update said in February.
There is also a lack of comprehensive regulations governing the use of FDP in criminal investigations when compared to other forms of biometrics.
“As this technology continues to evolve, ongoing dialogue among scientists, ethicists, legal experts, and the public will be essential to navigate the complex landscape of forensic genetics responsibly,” Biometric Update said in February. “…its application must be tempered with rigorous scientific validation, ethical considerations, and appropriate legal frameworks to ensure that the pursuit of justice does not come at the expense of individual rights and societal values.”
Interestingly, this earlier sentiment was in line with the researchers who developed Difface.
“To mitigate these risks, we emphasize the urgent need for comprehensive interdisciplinary analysis. Establishing clear, enforceable ethical frameworks and guidelines is paramount to ensuring that genomic research advances responsibly, safeguarding individual privacy and rights while harnessing the potential benefits of this technology. Such frameworks will help balance scientific progress with societal responsibility, creating a foundation for ethical and transparent application of DNA phenotyping across various fields,” the study said in the ‘Discussion’ section.
DNA is not the only tool which can recreate individual’s faces however. Facial recognition templates – datasets which can verify a face yet do not contain all the face’s data – can now be exploited to recreate the original facial image, according to a recent study.
“We present the first practical score-based face reconstruction and impersonation attack against three commercial FRS APIs: AWS CompareFaces, FACE++, and KAIROS, as well as five commonly used pre-trained open-source FRSs,” the study said in the ‘Abstract’ section. “Our attack is carried out in the black-box FRS model, where the adversary has no knowledge of the FRS (underlying models, parameters, template databases, etc.), except for the ability to make a limited number of similarity score queries. Notably, the attack is straightforward to implement, requires no trial-and-error guessing, and uses a small number of nonadaptive score queries. We motivate the attack by analyzing the topological meaning of similarity scores and then present our novel method using orthogonal face sets: a precomputed approximate basis set of human-like face images that enables us to get meaningful similarity scores from a small number of non-adaptive queries. Our approach successfully reconstructs human-like impersonation images with >20% (resp. >96%) success rates across three test datasets when directly attacking the AWS CompareFaces API (resp. open-source CosFace FRS) using only 100 queries—up to two orders of magnitude fewer queries than previous approaches. We provide evidence that personally identifiable biometric features are captured in our reconstructions by evaluating our approach in transfer-like attack settings and through other image similarity metrics.”