BiometricUpdate, August 22, 2022
The face image quality standard, which could be in place by early 2024, is being elaborated by the US government. This vendor-independent standard will serve to evaluate the usefulness of an image for biometric identification. The Department Homeland Security (overseen by the TSA and CBP) and the National Institute for Standards and Technology are collaborating on this standard.
For example, according to Bloomberg, demonstrating proper lighting based on the standard can remove some of the image’s bias towards darker skin tones.
According to Idemia North America CEO Donnie Scott, the government has already successfully used NIST as a technology evaluator, namely the well-known Facial Recognition Vendor Test, a comprehensive and ongoing program of testing biometric algorithms for strengths and weaknesses, among which is demographic bias. Scott also added that NIST’s work in this direction needs to continue. In his opinion, compliance with NIST standards should be a prerequisite for contractors participating in tenders for public works.
NYTimes, August 21, 2022. Updated August 25, 2022
We are sure that everyone knows this story, but we cannot help covering it. So many people around the world use remote consultations with medical specialists, send photos, record short videos, and few people think about the consequences of these actions, as, in fact, happened to Mark.
Mark, a stay-at-home father from San Francisco, took photos of his son’s illness with his Android smartphone to document the problem and track its progress. These photos were forwarded to his wife, who in turn sent them to the doctor. The doctor made a diagnosis, prescribed treatment, the child recovered. Everything is harmless, at first glance. Nothing criminal. But it all turned into a police investigation, and the loss of more than ten years of contacts, emails, and photos. Mark is caught in an algorithmic network designed to catch people sharing child sexual abuse material.
NIST has published its second draft of the AI RMF, encourages profile contributions and written comments by September 29, 2022. Discussion and comments are also welcome on October 18-19, 2022 at the NIST AI RMF workshop #3.
NIST is developing a risk management framework for various entities, be it an individual, society, or organization, in one way or another connected with AI. It is being developed through a collaborative open and transparent process. And our team, Adversa AI, has also contributed to this.
This framework is voluntary and serves to enhance the ability to incorporate reliability considerations into the design, development, use, and evaluation of AI products, services, and systems.
According to the world’s trends and the frequency of discussion of AI regulations addressed at a governmental level, it is clear that a wave of frameworks and regulations is coming, such as the DOE AI Risk Management Playbook (AIRMP).
Stay up to date with what is happening! Get a first look at news, noteworthy research and worst attacks on AI delivered right in your inbox.
Carrying out vulnerabilities in machine learning models as part of the study is necessary for further successful work on potential attacks and defenses. And here is a selection of the ...
