15 November 2024, Fri |
1:46 AM
By Vito Fabbrizio – Managing Director Biometric BU at HID Global
With recent advances in artificial intelligence, facial recognition is emerging as a powerful means of validating identity. People use it all the time to access their mobile phones, but acceptance hasn’t always extended beyond that personal use.
That’s due in part to concerns about potential misuse. Some people see Orwellian undertones in the notion of scores of databases full of facial scans. This points to the importance of facial recognition ethics, including discussions on how to measure and resolve disparities.
Biometric technologies are advancing rapidly, which means new modes of authenticating people are becoming more prevalent. Biometric facial recognition technology transforms images into numerical expressions, and computer algorithms make it possible to compare two images to see if they match.
The appropriate use of facial recognition technology depends on the prevailing culture, ethics, legislation and practices. With facial recognition still relatively untried — with no widely used or accepted regulations governing its use — security leaders will need to ensure they’re using the technology responsibly.
And they need to get moving on this soon. The market for facial recognition technology is expected to top $12.6 billion by 2028, up from $5 billion in 2021.
As a form of biometric authentication, facial recognition depends on artificial intelligence (AI) to identify human faces in images or videos. This approach can yield several key advantages including:
• Proof of presence — know for sure who accessed what and when
• Nothing to carry or remember — ATM cards, PINs and passwords potentially become obsolete
• Reduced human intervention — facial recognition takes the burden off human operators and steers toward contactless processes in the post-pandemic environment
AI plays a vital role in all of this, driving the high-volume data operations needed to scan and match faces at scale. This in and of itself has raised concern among privacy advocates and others who worry that automated algorithmic approaches to identity could potentially be misused.
Bias is an inherent human trait. Bias can be reflected and embedded in everything we create, even technology. In a recent World Economic Forum article, these biases were described as outputs that emanate from societal biases and include race, gender, biological sex, nationality or age.
How do such biases wind up in AI technology? AI algorithms must be trained by humans who use potentially unrepresentative or incomplete data that reflect historical inequalities. This can lead to biased algorithms and in turn, biased decisions that have a collective impact on certain groups of people. The Pew Research Center released data on facial recognition showing that only 5% of Americans “have a great deal of trust that technology companies will use facial recognition responsibly.”
Taking steps toward more ethical facial recognition technologies means addressing bias in the first place. The National Institute of Standards and Technology (NIST) recommends widening the scope of where we look for the source of biases, including going “…beyond the machine learning processes and data used to train AI software to the broader societal factors that influence how technology is developed.”
Other interesting ethical questions can arise at the intersection of facial recognition and business strategy. Researchers ask, for example, “if you use facial recognition to identify people coming into a store, should you use that identity to pull purchasing history? How about a credit report?”
Clearly, a thoughtful approach to facial recognition is needed going forward.
While the industry at large works to address the ethics in AI, organizations can adopt a framework for the ethical use of facial recognition technology. The American Civil Liberties Union submitted An Ethical Framework for Facial Recognition to the U.S. Department of Commerce’s National Telecommunications and Information Administration outlining several key best practices that include principles around the collection, use, sharing and access.
What does this mean for organizations who want to utilize facial recognition technology? It will be critical to outline and communicate:
• Informed consent — Those looking to implement a facial recognition solution need to consider “when and how to provide meaningful notice and to obtain their informed consent, especially if those individuals are then identified or profiled against other datasets,” according to the Future of Privacy Forum (FPF)
• Transparency tools — “Transparency has been suggested as one enabler to trust,” according to researchers at the Wilson Center, a non-partisan policy forum. They point to approaches that allow visibility into the inner workings of technical systems, such as explainable AI, open data and open algorithms. “Other strategies focus on exploring the outputs of an algorithm, including through testing that evaluates risks such as bias.”
• Privacy/ownership of data — It will be important, too, to establish rulers of the road spelling out privacy protections for consumers, and to develop a framework for ownership of facial recognition data. “Basic privacy principles require individuals to be aware of commercial entities that collect data about them with facial recognition systems, that they can request to know what data has been maintained on them and to request access to correct errors or delete information,” according to FPF.
• Governance — Organizations looking to implement facial recognition will need to establish clear governance. It will be important that they develop purposeful boundaries. They’ll need to determine and document the intended uses of facial recognition data and systems, and will need to have protocols in place that restrict the use of those solutions to only those predefined purposes.
Facial recognition technology has infinite potential to address real-world needs, including those of consumers and citizens. From retail settings to IT engagements to physical access control, facial recognition biometrics are in use today and will continue reshaping how we view identity.
However, addressing ethical AI and the concerns around facial recognition technology will be critical.
While the main problem of facial recognition technology stems from the lack of diversity in datasets, adopting ethical principles can help organizations mitigate risk and alleviate concerns when in use.
