Facial Recognition in Anti-Corruption

Facial recognition technology, which uses computer vision algorithms to identify or verify an individual's identity from a digital image or video of their face, has been applied to a range of anti-corruption tasks including monitoring restricted access, detecting fraudulent public payrolls, deterring bid rigging, tracking sanctioned individuals' cross-border travel, and tracing concealed assets. The APEC Policy Support Unit's Issues Paper No. 15 documents these applications within the emerging technology tier of the anti-corruption value chain, while giving substantial weight to the technology's well-documented error rates and privacy implications. ^[1]

1 Applications

The paper identifies five distinct anti-corruption applications of facial recognition technology. The first is monitoring access to detect unauthorized meetings between public officials and private parties, for example flagging when an official and a regulated party under their supervision are recorded entering the same restricted facility, a pattern that might otherwise go unnoticed. The second is ghost employee detection, verifying that individuals recorded as receiving a government salary or benefit correspond to real, physically present personnel, addressing a long-standing form of payroll fraud in which salaries continue to be paid against fabricated or duplicated employee identities. The third is bid rigging prevention, using facial recognition at tender or bid submission events to detect suspicious patterns of attendance, such as the same individuals repeatedly appearing as representatives of nominally competing firms, a signal of possible collusion. The fourth is travel monitoring, cross-referencing facial recognition data at borders or transit hubs to detect sanctioned individuals attempting to evade travel bans imposed as part of corruption-related sanctions. The fifth is asset tracing, cross-referencing security camera footage and photographs from real estate listings against public records and known images of a person of interest, to help expose unexplained wealth inconsistent with a public official's declared income (see Asset Tracing and Recovery Technology). ^[1]

2 INTERPOL's Facial Recognition System

The paper's principal institutional example is INTERPOL's Facial Recognition System (IFRS), launched in 2016. IFRS employs a dual-layer verification process, combining an automated software match against INTERPOL's image database with a mandatory human officer review of any candidate match, and classifies each search result into one of three categories: "Potential candidate," "No candidate," or "Inconclusive." ^[2] IFRS is used in border control screening and in processing INTERPOL Notices and Diffusions, the international alerts used to request the location, detention, or extradition of wanted individuals. The system requires that submitted images meet ICAO passport-standard image quality, the same technical specification used for machine-readable passport photographs, in order to be processed reliably. ^[2] INTERPOL's Face Expert Working Group, the body overseeing the system's technical standards, partners with the World Economic Forum and the United Nations Interregional Crime and Justice Research Institute (UNICRI) on governance and best-practice questions surrounding the technology's use. ^[2]

3 Documented error rates

The paper presents a significant documented risk associated with facial recognition deployment in law enforcement contexts more broadly, which bears directly on its reliability for anti-corruption use. Data from South Wales Police, following a facial recognition deployment beginning in 2017, showed that more than 89 percent of facial recognition flags generated by the system were incorrect. ^[3] The paper also notes that the use of facial recognition by London police has drawn sustained controversy from civil society organizations concerned about accuracy, oversight, and the absence of a clear legal framework governing its use. ^[3] These findings are cited as a caution against treating facial recognition matches, whether in anti-corruption or general law enforcement use, as reliable standalone evidence rather than a lead requiring independent human verification and corroboration.

4 Privacy and due process concerns

Beyond raw error rates, the paper identifies broader privacy and due process concerns attaching to facial recognition use. These include surveillance conducted without the knowledge or consent of the individuals monitored, materially higher error rates for women and people of color that have been documented across multiple facial recognition systems and vendors, and the risk posed by breaches of the biometric data such systems collect and store, since a face, unlike a password, cannot be changed after it is compromised. ^[1] These concerns parallel, on a technological axis, the procedural safeguards addressed in Procedural Fairness and reinforce the paper's broader theme that emerging surveillance-adjacent technologies carry governance costs that must be weighed against their detection benefits before wide deployment.

See also: Anti-Corruption Value Chain, Asset Tracing and Recovery Technology.

References

1. APEC Policy Support Unit, Technologies for Preventing, Detecting, and Combating Corruption, Issues Paper No. 15, June 2025, APEC#225-SE-01.10, section on facial recognition.
2. Ibid., section on INTERPOL's Facial Recognition System (IFRS), launched 2016, and the INTERPOL Face Expert Working Group.
3. Ibid., citing South Wales Police facial recognition deployment data since 2017 and controversy over London police use.