Among the aid that the world is bringing to Ukraine, to help it fight the war against Russia, there is also artificial intelligence. Clearview, an American company that provides a facial recognition platform based on intelligent biometric analysis that transposes information from 3D photographs, has offered its system to the Ukrainian government. Its clients include the Federal Bureau of Investigation, Immigration and Customs Enforcement, and the Fish and Wildlife Service, which use this tool to investigate and resolve crimes. It is one of the largest companies in the world in this industry, it is based on 10 billion images which have been retrieved from the Internet and in particular from sites such as Facebook, Instagram, Flickr and Getty Images. And as for the current situation, it also has over 2 billion portraits that come from the main Russian social media VKontakte. It is its strong point over another platform, PimEyes. Of Polish origin, the latter has often been used in conflicts and is open access, which means that anyone can go looking for photos and identify people. However, it is much more limited than Clearview.
The offer was made soon after the invasion began, but only now has the Ukrainian defense ministry revealed that it accepted it. Clearview also specified that it had not made any similar offers to Russia.
What has been made available is in practice a search engine which allows the Ukrainian authorities to filter people who show up at checkpoints. Not only. Artificial intelligence-based technology is supposed to serve the military for identify infiltrators and Russian militaryrefugees who have been separated from their families, but also unmask fake posts that spread disinformation and name the victims. It databases allows you to recognize the dead even if there is damage to the faceit also manages to interpret age-related changes and is more efficient than taking fingerprints.
Facial recognition will undoubtedly help, though many experts argue that it still has many limitations. In fact, it does not always provide correct results. The fear is that it could lead to misidentifications and therefore to the punishment of innocent people, or to the attribution of loss of life to people who are still around.
The American National Institute of Standards and Technology (NIST) has done a study on its effectiveness and concluded that it is not very accurate when it comes to identifying non-Caucasian faces. The worst results, with errors of 10 to 100 percent, are obtained with Asians, African Americans and indigenous people such as American Indians, and Pacific Islanders. But a quarter of the Russians live on the Asian side, with inhabitants, such as Tatars, Kazakhs and Mongols, who have characteristics typical of these populations.
Who is making efforts to improve it and make it independent of ethnic groups is there China, the country that has developed facial recognition the most. In cities, cameras that identify hundreds of people attending shops and offices are constantly increasing, 1.4 billion people per second.
However, Chinese companies are also moving to Africa. A contract has been signed with the Zimbabwean government. Beyond the interests that this country has towards the African continent, in this way they hope to obtain a much broader database. The difference in sampled faces improves the algorithm and then allows it to be sold elsewhere.
The fact remains that these types of services pose privacy concerns. Mass checks are in theory illegal. In many countries, including England and the USA, Clearview was subject to fines. Also in Italy the Privacy Guarantor, in February, awarded the company a fine of 20 million euros for having stored images of Italians, and asked that they be deleted. The accusation is of not having conducted operations in transparency, not having informed users and having used the data for purposes other than those that had been published online.
Obviously, in the event of a conflict, no one raises this problem. Furthermore, to avoid mistakes, the recognition should be used at the same time as other systems, but given the emergency situation it is difficult for this to happen.
The conflict
Online investigations and facial recognition against war crimes
by Federico Guerrini
