Dilemmas of data sharing in facial recognition

Facial recognition technology is everywhere these days. It’s in Facebook. It’s in mobile apps. It’s in ATMs. While it’s acceptance or at least proliferation in consumer technology grows, it’s use in the public sphere for public safety has faced bigger challenges.

Hollywood has set the bar high for facial recognition technology and what people expect. These expectations have been a challenge for the technology; however, the use of facial recognition technology is helping enhance public safety and assisting law enforcement agencies in identifying persons of interest.

Research shows that revenue for the global facial recognition market is forecast to reach US$882 million by 2024 at a CAGR of 22 percent, according to Tractica. High-profile terrorist attacks are one driving forces behind the rise in adoption, which has highlighted the need to strengthen security in public areas. In fact, Germany’s Minister of the Interior Thomas de Maiziere recently voiced support for using facial recognition in public spaces throughout Germany, such as airports and train stations.

Even with all the possible benefits, there have been many roadblocks for widespread adoption. From a lack of quality databases to technology shortcomings, the use of facial recognition t echnology particularly in public spaces has faced setbacks.

Sharing data for better security
Since an integral part of facial recognition technology relies on personal data, the sharing of data between different agencies and those that use the technology is crucial. However, sharing data isn’t always so easy. “Privacy laws differ from country to country, from state to state, and from company to company, often preventing the transfer of person data,” Elke Oberg, Marketing Manager at Cognitec Systems said. “Systems and standards also vary greatly, often preventing interoperability of image databases and metadata.”

Brian Lovell, CTO of Imagus Technology also pointed to legal issues being an obstacle for sharing data rather than a technical one. “Different agencies have access to different face databases for various purposes,” he said. “For example the driver’s license database used by transport is often not routinely available to police. However, with clever technology it is possible to search multiple databases in parallel without sharing the databases themselves.”

The public itself also stands in the way of data sharing. “It is not clear if the public will accept data sharing for the surveillance of public spaces, since people tend to demand a higher level of privacy in common areas,” Oberg added.

Despite these issues, there are ways suppliers of facial recognition technology could help facilitate the sharing of data. “Suppliers of facial recognition technology need to build full management and auditing capabilities into their systems for integrity reasons before court rulings mandate it,” Roger Rodriguez, Manager of Image Analytics at Vigilant Solutions opined. “Auditing needs to cover the facial recognition process from start to finish. This includes image capture, establishing image retention periods, and tracking all image deletions from the system. To supplement the management and auditing capabilities, vendors need to build output reporting mechanisms which cover system performance, quality assurance, and all system end-user audit trails.”

Rodriguez added, “Sharing images between agencies must also include a memorandum of understanding (MOUs) which sets protocols on usage rules and cover any agreed upon limits on data and image sharing. A supplier should integrate an electronic mechanism which allows agencies to retrieve these documents and store them within the systems. This will ensure full integrity between the sharing agencies and the supplier of the facial recognition system.”

More education on data collection
Since many in public safety analyze facial recognition technology based on success rate, most agencies that employ facial recognition focus on false positive rates and false negative rates. It is important, however, to consider other variables when measuring performance and establishing metrics. According to Rodriguez, “The quality of a probe image and the size of the gallery you are searching against always impact the data returned in each search. Smaller gallery sizes tend to yield higher accuracy results. Searching larger galleries means more faces to search against in a gallery. So a gallery sized in the millions is like looking for a needle in a haystack.”

By using filters like gender, race, age, height, and weight during searching helps to streamline the facial recognition process. “The key to data collection and analysis in facial recognition is educating these factors to users and administrators in the public safety space,” Rodriguez added. “This type of evangelism is needed to ensure higher accuracy rates are met during searches. This will also contribute to the integrity of data collection and the correct data analysis of facial recognition results.”

Facing the future of facial recognition
The unfortunate reality of large-scale attacks in crowded areas has stressed the need for enhanced public safety more than ever — facial recognition technology could help provide an extra layer of security. As technology advances and companies work hard to deal with the challenges head on, more widespread use could be around the corner.