ICE Facial Recognition Systems, Database Sources, and Demographic Error Rates

Immigration and Customs Enforcement (ICE) relies on smartphone tools to photograph and identify people in public spaces. Field agents from ICE and Customs and Border Protection (CBP) use an application named Mobile Fortify to scan faces during street encounters. That mobile app scans a subject's face against federal biometric repositories and returns personal biographic records. The software relies on matching technology developed by NEC. It represents one of two primary facial recognition systems tied to ICE operations.

The second system is SmartLINK. That app manages people on supervised release through required biometric check-ins. Together, Mobile Fortify and SmartLINK expand immigration surveillance from government checkpoints into public streets and private residences. Neither system published an independent accuracy audit. However, federal demographic testing on similar algorithms reveals elevated error rates for people of color. Understanding the reach of ICE facial recognition requires examining the software, the databases it queries, and the legal questions surrounding field scans.

Deployment of Mobile Fortify for Field Encounters

ICE facial recognition in field operations relies primarily on an application called Mobile Fortify. Agents run Mobile Fortify on handheld government smartphones during field encounters, traffic stops, and workplace actions. An agent points a phone camera directly at an individual to take a photograph. Mobile Fortify then transmits the captured picture to central biometric repositories run by the Department of Homeland Security (DHS).

The system returns biographic details within moments. It shows names, birth dates, and immigration history. The matching technology behind the software is built by NEC. DHS documented this operational deployment formally in its 2025 AI Use Case Inventory, which became public on January 29, 2026.

The deployment occurred without general public notice. Many individuals first encountered the tool during unscheduled stops. While airport terminals have long used biometric kiosks, field deployment puts biometric tracking on local sidewalks. Readers can compare this mobile system to border installations in Ban the Bots' overview of airport facial recognition.

Operational Mechanics and Biometric Database Checks

Mobile Fortify compares facial images against hundreds of millions of federal and state records. According to the ACLU of Wisconsin, the app checks faces against databases holding roughly 200 million images. These collections draw from passport records, visa applications, and state driver's license systems. The app also queries criminal-justice databases.

A single photo search reaches across distinct record systems. An agent does not need to know a subject's name beforehand. Instead, NEC algorithms calculate facial measurements and scan through federal repositories for close visual matches.

Retention policies raise serious privacy questions. DHS has stated it may retain photos taken through Mobile Fortify for up to 15 years. That retention applies regardless of a person's immigration status. It also applies if the photo produces no match at all. The data remains stored regardless of outcome.

This long retention window builds a permanent repository of field scans. Even people never accused of an immigration violation remain indexed in DHS archives. You can review general biometric storage concerns in Ban the Bots' guide to facial recognition technology.

Documented Public Encounters and Lawsuit Disclosures

Public disclosures reveal that Mobile Fortify is used heavily across the United States. According to a complaint filed by the Illinois Attorney General, ICE and CBP agents have used the app more than 100,000 times since its debut in 2025. That volume shows routine daily deployment by federal officers.

Detailed civilian encounters were documented in reporting by NBC News. In December 2025, an American citizen named Mubashir Khalif Hussen was detained by federal agents in Minneapolis after he refused to permit agents to scan his face. The detention followed his refusal to submit to a camera scan on a public street.

Other incidents in Minnesota followed a similar pattern. In January 2026, a Minneapolis mother named Katie Henly had her license plate and face photographed through her car window by a CBP agent. Her experience connects automated photo scans with mobile vehicle tracking, detailed in Ban the Bots' explainer on automated license plate readers.

A third incident involved a Minneapolis woman identified in news reporting only as Skye. Skye was detained and photographed by an ICE agent without explanation. The agent offered no legal justification for capturing her biometric image. These encounters demonstrate how field scans occur during everyday civilian activities. The encounters led to community protests and formal legal challenges against DHS tactics.

SmartLINK Facial Recognition under ISAP Monitoring

Outside of street encounters, ICE uses facial recognition to track people enrolled in release programs. The agency's Intensive Supervision Appearance Program (ISAP) is its primary alternative to physical custody. ISAP uses an application named SmartLINK to monitor participants. This software turns an enrollee's personal smartphone into a mandatory supervision checkpoint.

SmartLINK requires participants to complete scheduled check-ins using facial-recognition selfies. A participant uploads a photo. The software compares the image against a stored profile picture to confirm identity. Private government contractor BI, Incorporated manages ISAP and the app. BI, Incorporated operates as a wholly owned subsidiary of the private prison firm the GEO Group, as documented by the American Immigration Council.

Enrollment in this digital tracking network expanded rapidly. As of March 25, 2023, the American Immigration Council reported 281,613 people enrolled in the Alternatives to Detention program:

• 252,185 people were monitored through the SmartLINK smartphone app.
• 4,874 people were monitored via GPS ankle monitors.
• 12,522 people checked in through telephonic reporting systems.
• 12,032 people had no technology-based monitoring conditions.

SmartLINK covered more enrollees than the other three monitoring methods combined. Missing a biometric check-in can lead to re-arrest or revoked release. When the app fails to authenticate a face due to poor lighting or camera errors, the participant bears the burden of proof. This automated supervision operates without human review during the initial image capture.

Centralization under DHS Homeland Advanced Recognition Technology

DHS is working to replace its older biometric database, IDENT, with a centralized architecture called the Homeland Advanced Recognition Technology system (HART). IDENT served for decades as the central repository for fingerprints and immigration mugshots. HART is designed to combine face recognition with fingerprints, iris scans, and other biometric and biographic data in a single, centralized database.

The centralization of biometric files creates major civil liberties concerns. Civil liberties researchers at the Electronic Frontier Foundation (EFF) warned that HART is designed to combine distinct records to let DHS map non-obvious relationships between people. Bringing photographs, iris scans, and biographic files into one system allows cross-referencing on an unprecedented scale.

HART remains under active construction by DHS contractors. The department has not finalized an official completion date for the migration. But when completed, HART will store records collected from airports, border checkpoints, and mobile applications like Mobile Fortify. Storing multiple biometric markers in a central system increases the consequences of any misidentification.

Demographic Bias Research and Accuracy Disparities

Neither Mobile Fortify nor SmartLINK has published an independent, third-party accuracy audit. The public does not know the exact false-positive rate for either tool. However, extensive federal research demonstrates that facial recognition algorithms consistently show higher error rates for people of color.

The benchmark federal study came from the National Institute of Standards and Technology (NIST). In December 2019, NIST published its Face Recognition Vendor Test Part 3: Demographic Effects (NISTIR 8280). NIST researchers evaluated 189 algorithms from 99 commercial developers using 18.27 million images of 8.49 million individuals drawn from State Department, DHS, and FBI databases.

The test results were striking. For one-to-one matching, false-positive rates were 10 to 100 times higher for Asian and African American faces than for white faces, depending on the algorithm. That disparity means a person of color is far more likely to be matched to someone else's file. NIST also found that American Indian faces generated the highest false-positive rates among algorithms developed in the United States.

In one-to-many searches against an FBI database of 1.6 million mugshots, African American women showed elevated false-positive rates. These accuracy failures extend across commercial models. A 2018 study titled Gender Shades by MIT Media Lab researchers Joy Buolamwini and Timnit Gebru tested commercial gender-classification models from IBM, Microsoft, and Face++ across 1,270 subjects. The IBM system produced an error rate of 34.7% for darker-skinned women, compared to an error rate under 1% for lighter-skinned men.

The Gender Shades research evaluated gender classification rather than identity matching. Yet it demonstrated that computer vision models struggle on darker skin tones. When an officer uses mobile software on the street, demographic error gaps create an immediate risk of wrongful interrogation. Readers can study algorithmic disparities in Ban the Bots' guide to facial recognition ethics.

Systemic Skew in Underlying Law Enforcement Databases

Biometric errors do not stem solely from camera sensors or algorithmic code. The composition of the underlying image repositories also shapes who gets flagged. In 2016, the Georgetown Law Center on Privacy & Technology published a landmark report titled The Perpetual Line-Up.

The report revealed that more than half of all American adults were enrolled in a law-enforcement-searchable face recognition network. Most Americans are indexed through state driver's license databases. In many states, police and federal agencies search DMV photos without a warrant or judicial oversight.

The Georgetown researchers emphasized that criminal-justice databases carry pre-existing racial disparities. Because Black Americans face documented disparities in policing and arrest rates, their images appear disproportionately in mugshot collections. When ICE queries criminal-justice databases alongside DMV and passport photos, those historical disparities carry directly into the results. A demographic group overrepresented in an arrest database faces a higher mathematical risk of false-positive matches during broad searches.

When combined with the demographic error rates documented by NIST, database skew compounds the danger for minority communities. An algorithm prone to misidentifying Black faces is searching against databases where Black faces are disproportionately indexed. This mathematical reality undermines claims that algorithmic matching provides an objective tool for immigration stops.

Legal Recourse and Practical Guidance for Encounters

Federal deployment of mobile biometric scanners operates in an unsettled legal area. Most state and federal courts have not established clear rules for smartphone photo captures during street stops. A few municipalities have banned municipal agencies from using face scanning, but federal agents from ICE and CBP do not follow local bans.

When stopped by federal agents demanding a photograph, individuals have constitutional protections. You have the right to remain silent. You can ask an agent whether you are free to leave or if you are being detained. If an agent states you are detained, you have the right to speak with an attorney before answering questions.

Documenting encounters remains critical. Witnesses have the legal right to record federal officers operating in public spaces, provided they do not physically obstruct the officers. Capturing badge numbers, agency names, and the phone devices used during a stop creates evidence for administrative complaints or court challenges. Those evaluating legal limits on surveillance can consult Ban the Bots' guide to facial recognition laws.

Advocacy groups and legal organizations continue to file litigation against federal biometric programs. The Illinois Attorney General complaint against Mobile Fortify challenges field scans conducted without reasonable suspicion. To track ongoing legal challenges and court filings, visit Ban the Bots' tracker on AI lawsuits and practical strategies for fighting back against invasive surveillance.

Understanding your rights during street stops provides protection while courts address the constitutional limits of mobile ice facial recognition.

FAQ

Does ICE use facial recognition?

Yes, ICE uses facial recognition through multiple operational tools. Field agents use a smartphone application called Mobile Fortify to capture photos of individuals on the street and compare them against federal biometric databases. Separately, ICE monitors immigrants on supervised release through the SmartLINK smartphone app, which requires facial-recognition selfies during scheduled check-ins.

What is Mobile Fortify?

Mobile Fortify is a handheld smartphone application used by ICE and CBP officers during field encounters. An agent uses the phone camera to photograph a person's face, which the app matches against DHS biometric repositories using NEC algorithms. The app checks roughly 200 million images from passports, visas, driver's licenses, and criminal databases to return personal details.

What is SmartLINK and how does it use facial recognition?

SmartLINK is a mobile application used in ICE's Intensive Supervision Appearance Program, which serves as an alternative to immigration detention. Enrolled individuals must complete periodic check-ins by uploading a facial-recognition selfie. The software matches the selfie against a stored profile photo to verify that the participant is complying with supervision terms.

Is ICE's facial recognition technology accurate for people of color?

Neither Mobile Fortify nor SmartLINK has released independent accuracy audits. However, federal testing by NIST found that facial recognition algorithms generally produce false-positive rates 10 to 100 times higher for Asian and African American faces in one-to-one matches compared to white faces. Additionally, underlying mugshot databases disproportionately include Black individuals due to policing disparities.

What is the HART database?

HART stands for the Homeland Advanced Recognition Technology system, which DHS is building to replace its older IDENT biometric database. HART is designed to combine facial recognition, fingerprints, iris scans, and biographic records in a single centralized repository. Civil liberties groups have raised concerns that combining these datasets will allow DHS to map non-obvious relationships between people.