![Matt-Zbrog](https://www.forensicscolleges.com/wp-content/uploads/2018/12/Matt-Zbrog-150x150.jpg)
Matt Zbrog
Facial identification is one of the newer, sharper arrows in the forensic scientist’s quiver. Like fingerprint and DNA analysis, it draws from unique biological markers to identify suspects, victims, and persons of interest. In an increasingly recorded world, the applications of facial identification are vast, and AI and machine learning innovations are making the process quicker and more exact.
It can be a bedeviling subject to the untrained eye. A combination of factors contributes to making one face look like another or not. And for all the assistance AI and machine learning provide, they still need a trained human examiner to make the final call.
Facial identification is a relatively young forensics discipline that’s still changing as the associated technology, standards, and research evolve. Read on to learn more about facial identification, the challenges it must overcome, and where it’s headed in the future.
Lora Sims is director of the Face Center of Excellence (FaCE) and Biometrics SME at Ideal Innovations, Inc. (I-3), a company specializing in biometrics, forensics, and technical services. Starting with a background in tenprint, Sims has become a pioneer in developing and deploying internationally recognized facial identification training.
Currently chair of the Organization of Scientific Area Committees (OSAC) for Forensic Science Digital Multimedia Scientific Area Committee (DM SAC), Sims also serves as a member of the International Association for Identification’s (IAI) Science & Practices subcommittee for Facial Identification, a member and former chair of the Facial Identification Scientific Working Group (FISWG), and former chair of the OSAC Facial Identification subcommittee.
“The big challenge with faces is that they’re not static,” Sims says. “With fingerprints, the patterns are formed in utero and persist until after decomposition. Barring any scarring or disease, they’re permanent and unique. But faces are different: they change constantly.”
Faces change over time and under different conditions. How one ages, gains and loses weight, or affects different expressions will impact their face’s appearance. Different lighting, different poses, and even different clothes and accessories can complicate comparison further. We think we know a face—until we don’t.
“I started my career in fingerprints,” Sims says. “But I love the challenge of facial identification. It’s never cut and dry.”
Iris analysis is more static. But it’s also less widely applicable: an iris needs to be captured up close, similar to having someone take a brief eye exam. Many latent print examiners, Sims says, would prefer iris analysis to face analysis simply because it holds static features that are easier to identify and compare.
“Faces pose more challenges than irises or fingerprints,” Sims says. “Some people might think it’s easy to compare faces because we do it all the time with people we know. But it’s very different to compare unfamiliar faces.”
Research has repeatedly shown that trained facial examiners consistently outperform novices. That’s because there’s a meticulous science to the art. Facial examiners will analyze the features of the face for many things, including asymmetry, contour, and relative placement. They’ll then perform a comparative analysis against other images or database entries. And, unlike some automated AI-powered models, facial examiners are able to explain how they arrived at their conclusion in an objective and process-oriented manner, which is particularly valuable in the justice-oriented field of forensics.
“Facial recognition technology and systems have changed exponentially in the last decade,” Sims says. “With AI and deep neural networks, it’s incredible how fast and accurate they are. But they’re still just a tool to help humans. And that’s where we come in.”
Sims is the the director of the I-3 Face Center of Excellence (FaCE). When it was formed in 2015, FaCE offered only one class in facial identification training. Today, it’s expanded to eight different classes, with more in the pipeline. They offer 100 series, 200 series, 300 series, and 400 series courses.
“Our training classes are geared towards helping people learn how to do a comparison,” Sims says. “We teach people who have never compared anything before, and we also teach people who are already doing it and just need some additional training. Continuing education is important in every field, and it’s particularly important in forensics.”
The standardization of facial identification training is integral to the responsible development of the discipline. The field is still young, and forensic colleges don’t offer significant training in this area—yet. If and when they do, it’s pioneers like I-3 that curriculum designers will look to. It’s also where the corporate world, which has its own interest in facial identification, will look as well.
In late 2023, NEC, a multi-billion dollar electronics conglomerate, partnered with I-3 to implement forensic biometrics training. I-3 will provide training and operational support for customers using NEC’s facial recognition algorithms and systems in criminal investigations. While FaCE will continue to offer systems-agnostic training, the partnership has multiple benefits for the industry, chief among them being a large company like NEC instituting the best practice of making sure their users know how to do a competent, standards-based facial comparison.
Facial identification dates back hundreds of years, but in its modern instantiation, it’s still relatively young. The Facial Identification Scientific Working Group (FISWG) only started in 2009, and it’s focused on not only training practitioners doing comparisons but also the systems capturing and rendering facial images and standardizing both in ways that benefit the discipline as a whole. Meanwhile, research is paving the way forward for more efficient and responsible facial identification.
“One of the things being looked at is how humans handle the cross-race effect,” Sims says. “Does somebody of European ancestry perform well in facial identification only on subjects of European ancestry? How do they perform on subjects of African or Asian ancestry? That’s really important research.”
The future of facial identification will also be driven by further innovations in facial recognition technology. Advanced algorithms and improved AI should be able to assist more with complicating factors such as poor lighting, low resolution, or partial visibility. Integration with other types of biometric screening—fingerprints, iris scans, or even gait analysis—could contribute to multimodal forensic identification systems. But the need for qualified human examiners will only increase in the near term.
“Part of the responsible use of facial recognition technology is having people who are trained to adjudicate the candidate list,” Sims says. “Not just anybody can do that. People who are untrained do not do well with unfamiliar faces.”
Matt Zbrog
Matt Zbrog is a writer and researcher from Southern California. Since 2018, he’s written extensively about the increasing digitization of investigations, the growing importance of forensic science, and emerging areas of investigative practice like open source intelligence (OSINT) and blockchain forensics. His writing and research are focused on learning from those who know the subject best, including leaders and subject matter specialists from the Association of Certified Fraud Examiners (ACFE) and the American Academy of Forensic Science (AAFS). As part of the Big Employers in Forensics series, Matt has conducted detailed interviews with forensic experts at the ATF, DEA, FBI, and NCIS.