Facial recognition of senators

Use case ID: 068 
Author: Federal Senate of Brazil
Date: 3 October 2024

Objective:

Provide authentication of senators through facial recognition.

Actors:

Parliamentarians voting and registering their presence in Federal Senate sessions

Prerequisites:

Existing facial recognition model, trained with the faces of the senators in office

Integration with the Senado Digital app

Scenario:

Registering attendance

• A senator launches the Senado Digital app.
• The senator accesses the parliamentary area.
• The senator clicks to register attendance.
• The senator selects the site where the attendance will be registered.
• The senator selects the facial recognition login option.
• The app requests access to the device’s camera.
• The camera window opens, and the senator is instructed to position their face within the designated frame on the screen.
• If the senator’s face is successfully recognized, the app grants access to register attendance.

Registering a vote

• A senator launches the Senado Digital app.
• The senator accesses the parliamentary area.
• The senator clicks to register a vote in a committee.
• The senator selects the committee.
• The senator selects the facial recognition login option.
• The app requests access to the device’s camera.
• The camera window opens, and the senator is instructed to position their face within the designated frame on the screen.
• If the senator’s face is successfully recognized, the app grants access to register the vote.

Failed authentication

• A senator launches the Senado Digital app.
• The senator accesses the parliamentary area.
• The senator clicks to register attendance.
• The senator selects the site where the attendance will be registered.
• The senator selects the facial recognition authentication option.
• The app requests access to the device’s camera.
• The camera window opens, and the senator is instructed to position their face within the designated frame on the screen.
• The senator’s face is not recognized.
• The authentication options are presented again.

Expected results:

• Enhanced level of security, reducing the possibility of access and use of certain features by unauthorized individuals
• Identity fraud risk mitigation through the verification of unique biometric data, which is considerably harder to forge than passwords
• Reduced administrative burden, eliminating the need for password resets
• Improved user convenience, providing seamless authentication without the need to remember passwords

Potential challenges:

• Biases present in pre-trained models for detecting and generating face embeddings
• High latency during the facial recognition process
• Need to acquire enough facial samples from each parliamentarian to enable facial recognition with a high degree of accuracy and precision

Data requirements:

• Labelled samples of the faces of parliamentarians currently in office

Integrations with other systems:

• Senado Digital app
• Electronic Ballot Box Manager

Success metrics:

• Customer satisfaction scores
• Adoption rate (the proportion of authentications carried out using facial recognition since the release of the feature)
• Reduction in password resets