BBC Investigation: Moscow Authorities Employ Four Facial Recognition Algorithms

A recent BBC investigation has revealed that Moscow’s facial recognition system utilizes four different technologies, including one previously used in Belarus during the crackdown on the 2020 protests. In mid-summer, the Moscow city government’s subsidiary, AO “Electronic Moscow,” published four contracts for “data indexing” services within the city’s unified video analytics system. These contracts, awarded to a single supplier without a competitive bidding process or public documentation, total approximately 800 million rubles to be spent over two years.

According to a developer involved in one of the algorithms, these contracts essentially cover technical support and maintenance for the video analytics system. Each contract corresponds to a specific facial recognition algorithm, with the names listed on the government procurement website. The four algorithms are:

• NtechLab
• Tevian FaceSDK
• VisionLabs Luna Platform
• Kipod

Initially, Moscow’s system used three algorithms, with Kipod added later. BBC sources confirm that using multiple algorithms simultaneously is technically feasible. Artem Kozlyuk, head of the digital rights group Roskomsvoboda, suggests that employing four algorithms could increase the system’s effectiveness and reduce dependency on any single supplier, especially if a company faces sanctions or withdraws from cooperation.

Kipod: Belarusian Connection

Kipod was developed by the Russian-Belarusian company Synesis and became the main facial recognition technology used by Belarusian authorities to identify protest participants in Minsk in 2020. As a result, Synesis and one of its founders, Alexander Shatrov, were sanctioned by the EU and US. Synesis confirms that Kipod operates in Moscow’s metro stations. The technology can search for individuals not only by photo but also by characteristics such as approximate age, gender, race, and whether the person wears glasses, a beard, or a mustache.

VisionLabs Luna Platform

VisionLabs Luna Platform is used by several dozen Russian banks, including Alfa-Bank, to check potential clients against credit bureau blacklists before they even present documents. Other users include Sberbank, Post Bank, and Tinkoff. Luna, like Kipod, can determine facial parameters such as age, gender, race, and even “emotions,” according to user manuals. It is an official partner in Moscow’s Face Pay project, which allows metro passengers to upload a selfie, link a bank card, and access the metro with a glance at a camera.

Tevian FaceSDK and NtechLab

The fourth algorithm, FaceSDK, is developed by Moscow-based Tevian. All four companies are residents of the Skolkovo innovation center. Tevian claims FaceSDK can detect emotions such as anger, fear, sadness, joy, and surprise. According to Moscow’s Department of Information Technology, video streams from cameras are processed simultaneously by all four algorithms, and the final result is calculated using a special formula.

System Evaluation and Expansion

In early 2021, AO “Electronic Moscow” purchased a special program to automatically evaluate the performance of facial recognition algorithms. Developed by Amtech, a frequent winner of Moscow IT tenders, this system rates algorithms based on speed, accuracy, and efficiency. The procurement documents specify compatibility with FindFace, VisionLabs LUNA, Tevian FaceSDK, and Kipod. Nearly 800 million rubles will be spent on these algorithms over two years, with an additional 100 million rubles allocated for other components, such as a video detector for identifying human silhouettes.

Tevian and Synesis did not respond to BBC’s inquiries, while VisionLabs referred questions to the Department of Information Technology.

Data Collection and Surveillance in Moscow

The Moscow city government operates nearly 190 different information systems to collect and analyze data on citizens, buildings, vehicles, kiosks, playgrounds, and more. It also analyzes user behavior on the city’s Wi-Fi network and various municipal websites. For several years, work has been underway to create a unified system capable of extracting data from all city government resources and platforms, using machine learning to improve its performance.

Efforts to create a similar system at the federal level are documented in the “Network Freedoms” project’s report on surveillance technologies. The Ministry of Internal Affairs (MVD) attempted to develop the “IBD-F 2.0” system to replace the outdated “IBD-F,” which unifies police databases across Russia. The new system was intended to connect to other government databases, including public transportation and biometric data, but the contract was terminated due to “non-fulfillment of conditions.”

The MVD still operates other systems, such as the “Accounting and Barrier System — M” in Moscow, which allows cross-database searches and the creation of watchlists for individuals of operational interest. The city’s video surveillance system, integrated with facial recognition, assists in locating these individuals.

Use in Law Enforcement and Protests

According to OVD-Info, on June 12 alone, Moscow’s metro facial recognition system led to the detention of 35 people previously involved in protests. Human rights activists report that police told detainees they were flagged for “Russia Day.” Among those detained were Sota correspondent Petr Ivanov, Olga Bazhanova (previously charged for protest violations), science journalist Asya Kazantseva, and other activists.

Facial recognition is also being implemented in St. Petersburg, another city with a strong protest movement, leading journalists and rights advocates to expect similar incidents there. Roskomsvoboda continues to monitor detentions involving surveillance data. In November 2020, they reported on Sergey Mezhuyev, who was mistakenly identified by the metro’s facial recognition system. In January 2021, historian Kamil Galeev was detained a week after participating in a protest, tracked down using the system. Housing activist and Nabokov scholar Mikhail Shulman was also detained based on facial recognition results during a protest in January 2021. Municipal deputy Yulia Shcherbakova recounted being falsely accused and detained due to erroneous facial recognition results in July 2021.

Other victims include Vladimir Zalishchak, a municipal deputy from Moscow’s Donskoy district; Moscow City Duma deputy Elena Shuvalova; and, after the start of the so-called “special operation” in 2022, programmer Pavel, who protested on VK. Politician Leonid Gozman reported being detained three times via camera surveillance. In August 2022, Moscow’s Department of Transport reported identifying 221 people using these systems, mostly individuals wanted by police, though it’s unclear if opposition members are included in this number.

Data Leaks and Calls for Regulation

In summer 2020, Roskomsvoboda’s own investigation found that facial recognition systems are a constant source of data leaks. Volunteer Anna Kuznetsova was able to order a “trace” of her face through the cameras for a relatively small fee. The police officers who provided the data were held accountable, but the problem of leaks remains unresolved, as does the lack of public oversight over these systems.

Roskomsvoboda continues to advocate for a ban on facial recognition systems until regulatory mechanisms are developed that satisfy Russian society’s concerns about data use and privacy.