Live Facial Recognition Technology

What is the purpose of the live facial recognition trial?

The technology is a precision crime-fighting tactic aimed at improving public safety, reducing harm in our community, and preventing crime. It has been proven in other police jurisdictions to help police officers operate more efficiently to locate people they are lawfully trying to find.

The LFR trial aims to support police in finding and managing high-risk individuals within busy public spaces and identifying and locating people more quickly when needed.

All LFR deployments are conducted in accordance with WA Police Force Overt LFR Standard Operating Procedures.

Where is Live Facial Recognition (LFR) technology used?

LFR is used in intelligence-led and time-limited deployments. All deployments are overt and occur in public spaces where LFR has the greatest potential to assist the WA Police Force in fulfilling its operational duties.

Extra scrutiny is applied before authorising deployment near locations where there may be community sensitivity, such as hospitals, schools, and places of worship. An Equality Impact Assessment (PDF, 577KB) will guide police in mitigating impact with our community.

How will I know if an LFR deployment is in operation?

During deployments, overt awareness measures are used so the public can identify that LFR is in use and understand the nature of the operations.

These will include signage, posters with QR codes for links to further information. This webpage will provide notice ahead of deployments.

Who is on the alert-list used during an LFR deployment?

The alert-list consists of lawfully held images of people of interest to police, including individuals wanted for outstanding warrants, persons suspected of serious offences, individuals subject to lawful restrictions, missing persons, and persons who may pose a risk to themselves or others.

What happens if someone triggers an alert through LFR?

Every alert is checked by police officers before any action is taken. Considerations include how similar the images are, the similarity score, and environmental and technical factors. If appropriate, a police officer may speak with the individual to confirm their identity.  

An alert is not, by itself, grounds to arrest or exercise powers.  

Incorrect alerts are reviewed as part of post-deployment analysis.

What happens to images of people who are not on the alert-list?

Where no alert is generated, no information is stored; all derived data is immediately and automatically deleted.

Is live facial recognition legal?

Yes. LFR operates consistent with existing policing powers and the legal frameworks concerning taking images of people in public places. LFR is subject to privacy, equality, and data protection obligations.

Each deployment will be individually assessed to ensure it is lawful, necessary, and proportionate and then explicitly authorised. Authorisation addresses the policing purpose and ensures compliance with relevant legislation and privacy obligations.

How will the use of this technology be regulated?

Governance includes a rigorous process before, during and after each deployment. Each LFR deployment is endorsed by an Authorising Officer of the rank of Superintendent or above.

Before approving a deployment, the Authorising Officer must consider:

• whether the use is necessary and proportionate
• why the chosen location is appropriate
• what categories of people sought by police are to be included on the alert list
• all relevant assessments and safeguards

The process includes:

• strict approval before deployment
• command roles during deployment
• reporting and review afterwards

How is my privacy protected when LFR is used?

LFR is used in line with the WA Police Force Privacy Statement and Privacy Impact Statement (PDF, 571KB). Key protections include:

• where no alert is generated, the system automatically pixelates faces on the operator screen
• biometric data that does not generate an alert is not stored by police; it is deleted immediately and automatically
• biometric data is deleted following each deployment
• details of any alerts are retained, where required for an investigation of an offence or complaint
• no data is shared with third parties

What testing has been done to ensure the technology is accurate?

Police in the UK had the facial recognition algorithm independently performance tested by the UK National Physical Laboratory. It has also been evaluated by the National Institute of Standards and Technology. 

Testing shows high accuracy and minimal demographic variation in controlled environments. 

A conservative similarity score threshold is used by WA Police Force to minimise false alerts.

What is a similarity score?

This score measures how closely two facial patterns match. A higher score means the faces are more similar; a lower score means fewer similarities. Police set a pre-defined threshold so only strong matches trigger an alert.

What safeguards are in place to prevent bias and incorrect matches?

Safeguards include:

• conservative similarity threshold
• trained operators
• continuous monitoring
• human review of all alerts
• the ability to stop a deployment if appropriate
• evaluation reporting after every deployment

How will the LFR trial be evaluated?