7.3. Sensors

If device implementations include a particular sensor type that has a corresponding API for third-party developers, the device implementation MUST implement that API as described in the Android SDK documentation and the Android Open Source documentation on sensors.

Device implementations:

  • [C-0-1] MUST accurately report the presence or absence of sensors per the android.content.pm.PackageManager class.
  • [C-0-2] MUST return an accurate list of supported sensors via the SensorManager.getSensorList() and similar methods.
  • [C-0-3] MUST behave reasonably for all other sensor APIs (for example, by returning true or false as appropriate when applications attempt to register listeners, not calling sensor listeners when the corresponding sensors are not present; etc.).

If device implementations include a particular sensor type that has a corresponding API for third-party developers, they:

  • [C-1-1] MUST report all sensor measurements using the relevant International System of Units (metric) values for each sensor type as defined in the Android SDK documentation.

  • [C-1-2] MUST report sensor data with a maximum latency of 100 milliseconds + 2 * sample_time for the case of a sensor stream with a maximum requested latency of 0 ms when the application processor is active. This delay does not include any filtering delays.

  • [C-1-3] MUST report the first sensor sample within 400 milliseconds + 2 * sample_time of the sensor being activated. It is acceptable for this sample to have an accuracy of 0.

  • [SR] SHOULD report the event time in nanoseconds as defined in the Android SDK documentation, representing the time the event happened and synchronized with the SystemClock.elapsedRealtimeNano() clock. Existing and new Android devices are STRONGLY RECOMMENDED to meet these requirements so they will be able to upgrade to the future platform releases where this might become a REQUIRED component. The synchronization error SHOULD be below 100 milliseconds.

  • [C-1-4] For any API indicated by the Android SDK documentation to be a continuous sensor, device implementations MUST continuously provide periodic data samples that SHOULD have a jitter below 3%, where jitter is defined as the standard deviation of the difference of the reported timestamp values between consecutive events.

  • [C-1-5] MUST ensure that the sensor event stream MUST NOT prevent the device CPU from entering a suspend state or waking up from a suspend state.

  • When several sensors are activated, the power consumption SHOULD NOT exceed the sum of the individual sensor’s reported power consumption.

The list above is not comprehensive; the documented behavior of the Android SDK and the Android Open Source Documentations on sensors is to be considered authoritative.

Some sensor types are composite, meaning they can be derived from data provided by one or more other sensors. (Examples include the orientation sensor and the linear acceleration sensor.)

Device implementations:

  • SHOULD implement these sensor types, when they include the prerequisite physical sensors as described in sensor types.

If device implementations include a composite sensor, they:

  • [C-2-1] MUST implement the sensor as described in the Android Open Source documentation on composite sensors.

7.3.1. Accelerometer

Device implementations:

  • [C-SR] Are STRONGLY RECOMMENDED to include a 3-axis accelerometer.

If device implementations include a 3-axis accelerometer, they:

  • [C-1-1] MUST be able to report events up to a frequency of at least 50 Hz.
  • [C-1-2] MUST implement and report TYPE_ACCELEROMETER sensor.
  • [C-1-3] MUST comply with the Android sensor coordinate system as detailed in the Android APIs.
  • [C-1-4] MUST be capable of measuring from freefall up to four times the gravity(4g) or more on any axis.
  • [C-1-5] MUST have a resolution of at least 12-bits.
  • [C-1-6] MUST have a standard deviation no greater than 0.05 m/s^, where the standard deviation should be calculated on a per axis basis on samples collected over a period of at least 3 seconds at the fastest sampling rate.
  • [SR] are STRONGLY RECOMMENDED to implement the TYPE_SIGNIFICANT_MOTION composite sensor.
  • [SR] are STRONGLY RECOMMENDED to implement and report [TYPE_ACCELEROMETER_UNCALIBRATED] (https://developer.android.com/reference/android/hardware/Sensor.html#STRING_TYPE_ACCELEROMETER_UNCALIBRATED) sensor. Android devices are STRONGLY RECOMMENDED to meet this requirement so they will be able to upgrade to the future platform release where this might become REQUIRED.
  • SHOULD implement the TYPE_SIGNIFICANT_MOTION, TYPE_TILT_DETECTOR, TYPE_STEP_DETECTOR, TYPE_STEP_COUNTER composite sensors as described in the Android SDK document.
  • SHOULD report events up to at least 200 Hz.
  • SHOULD have a resolution of at least 16-bits.
  • SHOULD be calibrated while in use if the characteristics changes over the life cycle and compensated, and preserve the compensation parameters between device reboots.
  • SHOULD be temperature compensated.

If device implementations include a 3-axis accelerometer and any of the TYPE_SIGNIFICANT_MOTION, TYPE_TILT_DETECTOR, TYPE_STEP_DETECTOR, TYPE_STEP_COUNTER composite sensors are implemented:

  • [C-2-1] The sum of their power consumption MUST always be less than 4 mW.
  • SHOULD each be below 2 mW and 0.5 mW for when the device is in a dynamic or static condition.

If device implementations include a 3-axis accelerometer and a 3-axis gyroscope sensor, they:

  • [C-3-1] MUST implement the TYPE_GRAVITY and TYPE_LINEAR_ACCELERATION composite sensors.
  • [C-SR] Are STRONGLY RECOMMENDED to implement the TYPE_GAME_ROTATION_VECTOR composite sensor.

If device implementations include a 3-axis accelerometer, a 3-axis gyroscope sensor, and a magnetometer sensor, they:

  • [C-4-1] MUST implement a TYPE_ROTATION_VECTOR composite sensor.

7.3.2. Magnetometer

Device implementations:

  • [C-SR] Are STRONGLY RECOMMENDED to include a 3-axis magnetometer (compass).

If device implementations include a 3-axis magnetometer, they:

  • [C-1-1] MUST implement the TYPE_MAGNETIC_FIELD sensor.
  • [C-1-2] MUST be able to report events up to a frequency of at least 10 Hz and SHOULD report events up to at least 50 Hz.
  • [C-1-3] MUST comply with the Android sensor coordinate system as detailed in the Android APIs.
  • [C-1-4] MUST be capable of measuring between -900 µT and +900 µT on each axis before saturating.
  • [C-1-5] MUST have a hard iron offset value less than 700 µT and SHOULD have a value below 200 µT, by placing the magnetometer far from dynamic (current-induced) and static (magnet-induced) magnetic fields.
  • [C-1-6] MUST have a resolution equal or denser than 0.6 µT.
  • [C-1-7] MUST support online calibration and compensation of the hard iron bias, and preserve the compensation parameters between device reboots.
  • [C-1-8] MUST have the soft iron compensation applied—the calibration can be done either while in use or during the production of the device.
  • [C-1-9] MUST have a standard deviation, calculated on a per axis basis on samples collected over a period of at least 3 seconds at the fastest sampling rate, no greater than 1.5 µT; SHOULD have a standard deviation no greater than 0.5 µT.
  • SHOULD implement TYPE_MAGNETIC_FIELD_UNCALIBRATED sensor.
  • [SR] Existing and new Android devices are STRONGLY RECOMMENDED to implement the TYPE_MAGNETIC_FIELD_UNCALIBRATED sensor.

If device implementations include a 3-axis magnetometer, an accelerometer sensor, and a 3-axis gyroscope sensor, they:

  • [C-2-1] MUST implement a TYPE_ROTATION_VECTOR composite sensor.

If device implementations include a 3-axis magnetometer, an accelerometer, they:

  • MAY implement the TYPE_GEOMAGNETIC_ROTATION_VECTOR sensor.

If device implementations include a 3-axis magnetometer, an accelerometer and TYPE_GEOMAGNETIC_ROTATION_VECTOR sensor, they:

  • [C-3-1] MUST consume less than 10 mW.
  • SHOULD consume less than 3 mW when the sensor is registered for batch mode at 10 Hz.

7.3.3. GPS

Device implementations:

  • [C-SR] Are STRONGLY RECOMMENDED to include a GPS/GNSS receiver.

If device implementations include a GPS/GNSS receiver and report the capability to applications through the android.hardware.location.gps feature flag, they:

  • [C-1-1] MUST support location outputs at a rate of at least 1 Hz when requested via LocationManager#requestLocationUpdate.

  • [C-1-2] MUST be able to determine the location in open-sky conditions (strong signals, negligible multipath, HDOP < 2) within 10 seconds (fast time to first fix), when connected to a 0.5 Mbps or faster data speed internet connection. This requirement is typically met by the use of some form of Assisted or Predicted GPS/GNSS technique to minimize GPS/GNSS lock-on time (Assistance data includes Reference Time, Reference Location and Satellite Ephemeris/Clock).

    • [C-1-6] After making such a location calculation, device implementations MUST determine its location, in open sky, within 5 seconds, when location requests are restarted, up to an hour after the initial location calculation, even when the subsequent request is made without a data connection, and/or after a power cycle.
  • In open sky conditions after determining the location, while stationary or moving with less than 1 meter per second squared of acceleration:

    • [C-1-3] MUST be able to determine location within 20 meters, and speed within 0.5 meters per second, at least 95% of the time.
    • [C-1-4] MUST simultaneously track and report via GnssStatus.Callback at least 8 satellites from one constellation.
    • SHOULD be able to simultaneously track at least 24 satellites, from multiple constellations (e.g. GPS + at least one of Glonass, Beidou, Galileo).
  • [C-SR] Are STRONGLY RECOMMENDED to continue to deliver normal GPS/GNSS location outputs through GNSS Location Provider API's during an emergency phone call.

  • [C-SR] Are STRONGLY RECOMMENDED to report GNSS measurements from all constellations tracked (as reported in GnssStatus messages), with the exception of SBAS.

  • [C-SR] Are STRONGLY RECOMMENDED to report AGC, and Frequency of GNSS measurement.

  • [C-SR] Are STRONGLY RECOMMENDED to report all accuracy estimates (including Bearing, Speed, and Vertical) as part of each GPS/GNSS location.

  • [C-SR] Are STRONGLY RECOMMENDED to report GNSS measurements, as soon as they are found, even if a location calculated from GPS/GNSS is not yet reported.

  • [C-SR] Are STRONGLY RECOMMENDED to report GNSS pseudoranges and pseudorange rates, that, in open-sky conditions after determining the location, while stationary or moving with less than 0.2 meter per second squared of acceleration, are sufficient to calculate position within 20 meters, and speed within 0.2 meters per second, at least 95% of the time.

7.3.4. Gyroscope

Device implementations:

  • [C-SR] Are STRONGLY RECOMMENDED to include a gyroscope sensor unless a 3-axis accelerometer is also included.

If device implementations include a 3-axis gyroscope, they:

  • [C-1-1] MUST be able to report events up to a frequency of at least 50 Hz.
  • [C-1-2] MUST implement the TYPE_GYROSCOPE sensor and are STRONGLY RECOMMENDED to also implement the TYPE_GYROSCOPE_UNCALIBRATED sensor.
  • [C-1-4] MUST have a resolution of 12-bits or more and SHOULD have a resolution of 16-bits or more.
  • [C-1-5] MUST be temperature compensated.
  • [C-1-6] MUST be calibrated and compensated while in use, and preserve the compensation parameters between device reboots.
  • [C-1-7] MUST have a variance no greater than 1e-7 rad^2 / s^2 per Hz (variance per Hz, or rad^2 / s). The variance is allowed to vary with the sampling rate, but MUST be constrained by this value. In other words, if you measure the variance of the gyro at 1 Hz sampling rate it SHOULD be no greater than 1e-7 rad^2/s^2.
  • [SR] Calibration error is STRONGLY RECOMMENDED to be less than 0.01 rad/s when device is stationary at room temperature.
  • SHOULD report events up to at least 200 Hz.

If device implementations include a 3-axis gyroscope, an accelerometer sensor and a magnetometer sensor, they:

  • [C-2-1] MUST implement a TYPE_ROTATION_VECTOR composite sensor.

If device implementations include a 3-axis accelerometer and a 3-axis gyroscope sensor, they:

  • [C-3-1] MUST implement the TYPE_GRAVITY and TYPE_LINEAR_ACCELERATION composite sensors.
  • [C-SR] Are STRONGLY RECOMMENDED to implement the TYPE_GAME_ROTATION_VECTOR composite sensor.

7.3.5. Barometer

Device implementations:

  • [C-SR] Are STRONGLY RECOMMENDED to include a barometer (ambient air pressure sensor).

If device implementations include a barometer, they:

  • [C-1-1] MUST implement and report TYPE_PRESSURE sensor.
  • [C-1-2] MUST be able to deliver events at 5 Hz or greater.
  • [C-1-3] MUST be temperature compensated.
  • [SR] STRONGLY RECOMMENDED to be able to report pressure measurements in the range 300hPa to 1100hPa.
  • SHOULD have an absolute accuracy of 1hPa.
  • SHOULD have a relative accuracy of 0.12hPa over 20hPa range (equivalent to ~1m accuracy over ~200m change at sea level).

7.3.6. Thermometer

Device implementations:

  • MAY include an ambient thermometer (temperature sensor).
  • MAY but SHOULD NOT include a CPU temperature sensor.

If device implementations include an ambient thermometer (temperature sensor), they:

  • [C-1-1] MUST be defined as SENSOR_TYPE_AMBIENT_TEMPERATURE and MUST measure the ambient (room/vehicle cabin) temperature from where the user is interacting with the device in degrees Celsius.
  • [C-1-2] MUST be defined as SENSOR_TYPE_TEMPERATURE.
  • [C-1-3] MUST measure the temperature of the device CPU.
  • [C-1-4] MUST NOT measure any other temperature.

Note the SENSOR_TYPE_TEMPERATURE sensor type was deprecated in Android 4.0.

7.3.7. Photometer

  • Device implementations MAY include a photometer (ambient light sensor).

7.3.8. Proximity Sensor

  • Device implementations MAY include a proximity sensor.

If device implementations include a proximity sensor, they:

  • [C-1-1] MUST measure the proximity of an object in the same direction as the screen. That is, the proximity sensor MUST be oriented to detect objects close to the screen, as the primary intent of this sensor type is to detect a phone in use by the user. If device implementations include a proximity sensor with any other orientation, it MUST NOT be accessible through this API.
  • [C-1-2] MUST have 1-bit of accuracy or more.

7.3.9. High Fidelity Sensors

If device implementations include a set of higher quality sensors as defined in this section, and make available them to third-party apps, they:

  • [C-1-1] MUST identify the capability through the android.hardware.sensor.hifi_sensors feature flag.

If device implementations declare android.hardware.sensor.hifi_sensors, they:

  • [C-2-1] MUST have a TYPE_ACCELEROMETER sensor which:

    • MUST have a measurement range between at least -8g and +8g, SHOULD have a measurement range between at least -16g and +16g.
    • MUST have a measurement resolution of at least 2048 LSB/g.
    • MUST have a minimum measurement frequency of 12.5 Hz or lower.
    • MUST have a maximum measurement frequency of 400 Hz or higher; SHOULD support the SensorDirectChannel RATE_VERY_FAST.
    • MUST have a measurement noise not above 400 μg/√Hz.
    • MUST implement a non-wake-up form of this sensor with a buffering capability of at least 3000 sensor events.
    • MUST have a batching power consumption not worse than 3 mW.
    • [C-SR] Is STRONGLY RECOMMENDED to have 3dB measurement bandwidth of at least 80% of Nyquist frequency, and white noise spectrum within this bandwidth.
    • SHOULD have an acceleration random walk less than 30 μg √Hz tested at room temperature.
    • SHOULD have a bias change vs. temperature of ≤ +/- 1 mg/°C.
    • SHOULD have a best-fit line non-linearity of ≤ 0.5%, and sensitivity change vs. temperature of ≤ 0.03%/C°.
    • SHOULD have cross-axis sensitivity of < 2.5 % and variation of cross-axis sensitivity < 0.2% in device operation temperature range.
  • [C-2-2] MUST have a TYPE_ACCELEROMETER_UNCALIBRATED with the same quality requirements as TYPE_ACCELEROMETER.

  • [C-2-3] MUST have a TYPE_GYROSCOPE sensor which:

    • MUST have a measurement range between at least -1000 and +1000 dps.
    • MUST have a measurement resolution of at least 16 LSB/dps.
    • MUST have a minimum measurement frequency of 12.5 Hz or lower.
    • MUST have a maximum measurement frequency of 400 Hz or higher; SHOULD support the SensorDirectChannel RATE_VERY_FAST.
    • MUST have a measurement noise not above 0.014°/s/√Hz.
    • [C-SR] Is STRONGLY RECOMMENDED to have 3dB measurement bandwidth of at least 80% of Nyquist frequency, and white noise spectrum within this bandwidth.
    • SHOULD have a rate random walk less than 0.001 °/s √Hz tested at room temperature.
    • SHOULD have a bias change vs. temperature of ≤ +/- 0.05 °/ s / °C.
    • SHOULD have a sensitivity change vs. temperature of ≤ 0.02% / °C.
    • SHOULD have a best-fit line non-linearity of ≤ 0.2%.
    • SHOULD have a noise density of ≤ 0.007 °/s/√Hz.
    • SHOULD have calibration error less than 0.002 rad/s in temperature range 10 ~ 40 ℃ when device is stationary.
    • SHOULD have g-sensitivity less than 0.1°/s/g.
    • SHOULD have cross-axis sensitivity of < 4.0 % and cross-axis sensitivity variation < 0.3% in device operation temperature range.
  • [C-2-4] MUST have a TYPE_GYROSCOPE_UNCALIBRATED with the same quality requirements as TYPE_GYROSCOPE.

  • [C-2-5] MUST have a TYPE_GEOMAGNETIC_FIELD sensor which:

    • MUST have a measurement range between at least -900 and +900 μT.
    • MUST have a measurement resolution of at least 5 LSB/uT.
    • MUST have a minimum measurement frequency of 5 Hz or lower.
    • MUST have a maximum measurement frequency of 50 Hz or higher.
    • MUST have a measurement noise not above 0.5 uT.
  • [C-2-6] MUST have a TYPE_MAGNETIC_FIELD_UNCALIBRATED with the same quality requirements as TYPE_GEOMAGNETIC_FIELD and in addition:

    • MUST implement a non-wake-up form of this sensor with a buffering capability of at least 600 sensor events.
    • [C-SR] Is STRONGLY RECOMMENDED to have white noise spectrum from 1 Hz to at least 10 Hz when the report rate is 50 Hz or higher.
  • [C-2-7] MUST have a TYPE_PRESSURE sensor which:

    • MUST have a measurement range between at least 300 and 1100 hPa.
    • MUST have a measurement resolution of at least 80 LSB/hPa.
    • MUST have a minimum measurement frequency of 1 Hz or lower.
    • MUST have a maximum measurement frequency of 10 Hz or higher.
    • MUST have a measurement noise not above 2 Pa/√Hz.
    • MUST implement a non-wake-up form of this sensor with a buffering capability of at least 300 sensor events.
    • MUST have a batching power consumption not worse than 2 mW.
  • [C-2-8] MUST have a TYPE_GAME_ROTATION_VECTOR sensor.

  • [C-2-9] MUST have a TYPE_SIGNIFICANT_MOTION sensor which:

    • MUST have a power consumption not worse than 0.5 mW when device is static and 1.5 mW when device is moving.
  • [C-2-10] MUST have a TYPE_STEP_DETECTOR sensor which:

    • MUST implement a non-wake-up form of this sensor with a buffering capability of at least 100 sensor events.
    • MUST have a power consumption not worse than 0.5 mW when device is static and 1.5 mW when device is moving.
    • MUST have a batching power consumption not worse than 4 mW.
  • [C-2-11] MUST have a TYPE_STEP_COUNTER sensor which:

    • MUST have a power consumption not worse than 0.5 mW when device is static and 1.5 mW when device is moving.
  • [C-2-12] MUST have a TILT_DETECTOR sensor which:

    • MUST have a power consumption not worse than 0.5 mW when device is static and 1.5 mW when device is moving.
  • [C-2-13] The event timestamp of the same physical event reported by the Accelerometer, Gyroscope, and Magnetometer MUST be within 2.5 milliseconds of each other. The event timestamp of the same physical event reported by the Accelerometer and Gyroscope SHOULD be within 0.25 milliseconds of each other.

  • [C-2-14] MUST have Gyroscope sensor event timestamps on the same time base as the camera subsystem and within 1 milliseconds of error.

  • [C-2-15] MUST deliver samples to applications within 5 milliseconds from the time when the data is available on any of the above physical sensors to the application.

  • [C-2-16] MUST NOT have a power consumption higher than 0.5 mW when device is static and 2.0 mW when device is moving when any combination of the following sensors are enabled:

    • SENSOR_TYPE_SIGNIFICANT_MOTION
    • SENSOR_TYPE_STEP_DETECTOR
    • SENSOR_TYPE_STEP_COUNTER
    • SENSOR_TILT_DETECTORS
  • [C-2-17] MAY have a TYPE_PROXIMITY sensor, but if present MUST have a minimum buffer capability of 100 sensor events.

Note that all power consumption requirements in this section do not include the power consumption of the Application Processor. It is inclusive of the power drawn by the entire sensor chain—the sensor, any supporting circuitry, any dedicated sensor processing system, etc.

If device implementations include direct sensor support, they:

  • [C-3-1] MUST correctly declare support of direct channel types and direct report rates level through the isDirectChannelTypeSupported and getHighestDirectReportRateLevel API.
  • [C-3-2] MUST support at least one of the two sensor direct channel types for all sensors that declare support for sensor direct channel.
  • SHOULD support event reporting through sensor direct channel for primary sensor (non-wakeup variant) of the following types:
    • TYPE_ACCELEROMETER
    • TYPE_ACCELEROMETER_UNCALIBRATED
    • TYPE_GYROSCOPE
    • TYPE_GYROSCOPE_UNCALIBRATED
    • TYPE_MAGNETIC_FIELD
    • TYPE_MAGNETIC_FIELD_UNCALIBRATED

7.3.10. Biometric Sensors

For additional background on Measuring Biometric Unlock Security, please see Measuring Biometric Security documentation.

If device implementations include a secure lock screen, they:

  • SHOULD include a biometric sensor

Biometric sensors can be classified as Strong, Weak, or Convenience based on their spoof and imposter acceptance rates, and on the security of the biometric pipeline. This classification determines the capabilities the biometric sensor has to interface with the platform and with third-party applications. Sensors are classified as Convenience by default, and need to meet additional requirements as detailed below if they wish to be classified as either Weak or Strong. Both Weak and Strong biometrics get additional capabilities as detailed below.

To make a biometric sensor available to third-party applications, device implementations:

  • [C-0-1] MUST meet the requirements for Strong or Weak biometric as defined in this document.

To allow access to keystore keys to third-party applications, device implementations:

  • [C-0-2] MUST meet the requirements for Strong as defined in this document.

Additionally:

  • [C-0-3] MUST be paired with an explicit confirm action (e.g. a button press) if that Strong biometric is passive (e.g. face or iris where no explicit signal of the user's intent exists).
    • [C-SR] The confirm action for passive biometrics is STRONGLY RECOMMENDED to be secured such that an operating system or kernel compromise cannot spoof it. For example, this means that the confirm action based on a physical button is routed through an input-only general-purpose input/output (GPIO) pin of a secure element (SE) that cannot be driven by any other means than a physical button press.

If device implementations wish to treat a biometric sensor as Convenience, they:

  • [C-1-1] MUST have a false acceptance rate less than 0.002%.
  • [C-1-2] MUST disclose that this mode may be less secure than a strong PIN, pattern, or password and clearly enumerate the risks of enabling it, if the spoof and imposter acceptance rates are higher than 7%.
  • [C-1-3] MUST rate limit attempts for at least 30 seconds after five false trials for biometric verification - where a false trial is one with an adequate capture quality (BIOMETRIC_ACQUIRED_GOOD) that does not match an enrolled biometric.
  • [C-1-4] MUST prevent adding new biometrics without first establishing a chain of trust by having the user confirm existing or add a new device credential (PIN/pattern/password) that's secured by TEE; the Android Open Source Project implementation provides the mechanism in the framework to do so.
  • [C-1-5] MUST completely remove all identifiable biometric data for a user when the user's account is removed (including via a factory reset).
  • [C-1-6] MUST honor the individual flag for that biometric (i.e. DevicePolicyManager.KEYGUARD_DISABLE_FINGERPRINT, DevicePolicymanager.KEYGUARD_DISABLE_FACE , or DevicePolicymanager.KEYGUARD_DISABLE_IRIS ).
  • [C-1-7] MUST challenge the user for the recommended primary authentication (e.g. PIN, pattern, password) once every 24 hours or less for new devices launching with Android version 10, once every 72 hours or less for devices upgrading from earlier Android version.
  • [C-1-8] MUST challenge the user for the recommended primary authentication (eg: PIN, pattern, password) after one of the following:
    • a 4-hour idle timeout period, OR
    • 3 failed biometric authentication attempts.
    • The idle timeout period and the failed authentication count is reset after any successful confirmation of the device credentials.
  • [C-SR] Are STRONGLY RECOMMENDED to have a false rejection rate of less than 10%, as measured on the device.
  • [C-SR] Are STRONGLY RECOMMENDED to have a latency below 1 second, measured from when the biometric is detected, until the screen is unlocked, for each enrolled biometric.

If device implementations wish to treat a biometric sensor as Weak, they:

  • [C-2-1] MUST meet all requirements for Convenience above, except for [C-1-2].

  • [C-2-2] MUST have a spoof and imposter acceptance rate not higher than 20%.

  • [C-2-3] MUST have a hardware-backed keystore implementation, and perform the biometric matching in an isolated execution environment outside Android user or kernel space, such as the Trusted Execution Environment (TEE), or on a chip with a secure channel to the isolated execution environment.

  • [C-2-4] MUST have all identifiable data encrypted and cryptographically authenticated such that they cannot be acquired, read or altered outside of the isolated execution environment or a chip with a secure channel to the isolated execution environment as documented in the implementation guidelines on the Android Open Source Project site.

  • [C-2-5] For camera based biometrics, while biometric based authentication or enrollment is happening:

    • MUST operate the camera in a mode that prevents camera frames from being read or altered outside the isolated execution environment or a chip with a secure channel to the isolated execution environment.
    • For RGB single-camera solutions, the camera frames CAN be readable outside the isolated execution environment to support operations such as preview for enrollment, but MUST still NOT be alterable.
  • [C-2-6] MUST NOT enable third-party applications to distinguish between individual biometric enrollments.

  • [C-2-7] MUST NOT allow unencrypted access to identifiable biometric data or any data derived from it (such as embeddings) to the Application Processor outside the context of the TEE.

  • [C-2-8] MUST have a secure processing pipeline such that an operating system or kernel compromise cannot allow data to be directly injected to falsely authenticate as the user.

    If device implementations are already launched on an earlier Android version and cannot meet the requirement C-2-8 through a system software update, they MAY be exempted from the requirement.

If device implementations wish to treat a biometric sensor as Strong, they:

  • [C-3-1] MUST meet all the requirements of Weak above. Upgrading devices from an earlier Android version is not exempted from C-2-7.
  • [C-3-2] MUST have a spoof and imposter acceptance rate not higher than 7%.
  • [C-3-3] MUST challenge the user for the recommended primary authentication (e.g. PIN, pattern, password) once every 72 hours or less.

7.3.12. Pose Sensor

Device implementations:

  • MAY support pose sensor with 6 degrees of freedom.

If device implementations support pose sensor with 6 degrees of freedom, they:

  • [C-1-1] MUST implement and report TYPE_POSE_6DOF sensor.
  • [C-1-2] MUST be more accurate than the rotation vector alone.