Security About 2 min read

Privacy and sensitive data

Video, access and alarm records can contain sensitive information. Limit collection and access to the workflow's needs, and obtain the required privacy and governance review.

Sources and scopeSource record 25 August 2026

Technical source record: 25 August 2026. Check the linked documentation for current product requirements.

Research and static guidance only; product and deployment specific behaviour requires controlled environment validation and authoritative product evidence.

Verification and testing

Overview#

Physical security data can reveal identity, movement, associations, schedules, facility layout, access rights, behaviour, health related inferences, and response capability. Global technical guidance can't determine whether collection or use is lawful in a particular jurisdiction; obtain local legal and governance review.

Data classes#

Class Examples Special risk
Media Live/recorded video, audio, intercom calls Captures bystanders and sensitive spaces; secondary use is easy
Biometric Face, fingerprint, iris, voice template and quality data Difficult to replace; matching and demographic performance concerns
Credential Card/mobile identifiers, keys, access rights, revocation state Enables tracking, impersonation or access policy inference
Location and behaviour ANPR/LPR, UWB/BLE position, occupancy, access history Creates detailed movement and association records
Identity Names, photos, directory identifiers, visitor records Links system events to people and organizations
Security telemetry Alarms, camera health, blind spots, topology, response Can expose facility weaknesses and operational patterns

Data flow questions#

For every field ask: why is it needed, who is the subject, where is it collected, which systems receive it, which identifiers join it, how long is it retained, who can query/export it, which vendors/subprocessors handle it, what leaves the site/region, how is deletion propagated, and what happens if the purpose ends.

Engineering controls#

  • Collect the least resolution, duration, area, audio, biometric feature set, and metadata required for the declared purpose.
  • Apply privacy masks and exclusion zones at the earliest reliable stage, while documenting whether raw data exists upstream.
  • Separate operational viewing, investigation, export, analytics training, administration and bulk search permissions.
  • Default APIs to narrow time/site/object ranges; rate limit bulk enumeration and export.
  • Encrypt data in transit and at rest with separately managed keys and auditable access.
  • Redact logs and support bundles; use synthetic data in documentation and development.
  • Make retention and deletion explicit across edge storage, VMS, cloud replicas, caches, analytics stores, backups and exports.
  • Record model/version and evaluation limits for biometric or analytic decisions; keep human review where consequences warrant it.
  • Design for subject correction, revocation, deletion and account closure without corrupting audit integrity.

Developer anti patterns#

  • using a stable card number as a cross system analytics identifier without necessity;
  • placing faces, plate numbers or raw credential values in broker topic names or URLs;
  • returning unrestricted historical search results to a broad operator role;
  • retaining debug media or full API bodies indefinitely;
  • silently repurposing security footage to train analytics;
  • claiming anonymization while stable device, location and time fields permit re identification.

Sources#

Section overview · Wiki home