What are Technical and organisational measures?

Technical and organisational measures - definition

Technical and organisational measures (TOMs) are documented safeguards used to protect personal data during processing. Under Article 32 of the General Data Protection Regulation (GDPR), Regulation (EU) 2016/679, controllers and processors must implement measures appropriate to the risk. Article 32 requires consideration of the state of the art, implementation costs, the nature, scope, context, and purposes of processing, and the likelihood and severity of risks to individuals.

The GDPR does not prescribe one mandatory technology or security configuration. Instead, it requires a risk-based decision that can be justified, implemented, tested, and reviewed. Article 32(1) identifies pseudonymization and encryption as examples. Depending on the processing context, robust anonymization of image or video footage can also form part of a documented TOM framework. However, anonymization is not automatically achieved by applying a blur effect.

For image and video processing, TOMs address risks such as unauthorized access to identifiable footage, re-identification of individuals, inappropriate disclosure of footage, loss of source files, and inaccurate masking of faces or license plates.

How anonymization supports Article 32 GDPR compliance

Article 32 does not state that every face or license plate must be anonymized. It requires the controller and processor to select measures proportionate to the assessed risk. Anonymizing footage before it is shared with a vendor, used for training, retained for analysis, or exported outside a restricted workflow can reduce the impact of unauthorized disclosure.

Recital 26 GDPR distinguishes anonymous information from personal data. Information is anonymous only where the data subject is not identifiable by means reasonably likely to be used. If a blurred face can still be linked to a person through clothing, location, time, voice, captions, or associated records, the footage may remain personal data. A blurred video should therefore not be classified as anonymous without considering the full processing context.

A documented TOM decision for image and video anonymization should establish the following points:

  • the processing purpose and the categories of persons shown in the footage;
  • whether original, identifiable footage is necessary for that purpose;
  • the risk of identification from visual and contextual information;
  • the masking method, coverage requirements, and review process;
  • access restrictions for original and processed files;
  • retention periods, deletion procedures, and incident-response responsibilities.

Technical measures for face and license plate masking

Technical measures should protect both the original footage and the anonymized output. Face detection is used to locate faces in individual frames. A masking process then applies blur, pixelation, solid masking, or another transformation to the detected area across the relevant video sequence. License plate detection follows a comparable process for vehicle registration plates.

The following controls are commonly relevant when an organization processes recordings containing faces or license plates:

Control area

Example measure

Evidence for accountability

Access control

Role-based access to source footage and separate permissions for exports.

Access policy, authorization matrix, access-review records.

Data minimization

Process only the frames, files, and retain them only for the period needed for the defined purpose.

Processing specification and deletion schedule.

Masking quality

Review detection results and correct missed or incomplete masks before release.

Quality-assurance protocol and review records.

Confidentiality

Encrypt footage in transit and at rest and protect cryptographic keys.

Encryption standard, key-management procedure, system configuration.

Availability and resilience

Maintain backups, recovery procedures, and controlled processing environments.

Business continuity and restoration-test records.

Detection performance must be measured in the environment in which the system is used. Relevant indicators include recall, precision, false-negative rate, and the proportion of frames reviewed. Recall can be expressed as: recall = true positive detections / (true positive detections + false negative detections). A low false-negative rate is important because an undetected face or license plate can leave a person identifiable in the released footage.

Organisational measures and documented governance

Technical controls are not sufficient on their own. Organizational measures define who may process footage, when manual review is required, how errors are corrected, and how the organization responds to a security incident. These measures should be integrated with the records required by Article 30 GDPR, data protection impact assessments under Article 35 where applicable, processor agreements under Article 28, and breach-response procedures under Articles 33 and 34.

A practical governance process should include the following steps:

  1. Document the purpose, data categories, recipients, retention period, and risk assessment.
  2. Define acceptance criteria for face and license plate masking before footage is released.
  3. Test the workflow with footage representing expected lighting, camera angles, movement, occlusion, and image quality.
  4. Assign responsibility for quality checks, exceptions, export approval, and deletion.
  5. Review the TOMs after material changes to the footage source, processing method, threat environment, or intended use.

Application to Gallio PRO workflows

In a Gallio PRO workflow, automatic processing covers face detection and license plate detection for the purpose of applying masking. The software does not automatically detect company logos, tattoos, name badges, documents, or content shown on monitors. These elements can be masked manually with the built-in editor when they create an identification or confidentiality risk.

Gallio PRO is designed for processing files rather than real-time anonymization or video stream anonymization. A controller should define whether the original footage, the processed output, or both are retained, who can access each version, and when each version must be deleted. The organization should also verify that its processing configuration and operating procedures meet its own Article 32 risk assessment.

US equivalent

The United States has no single federal equivalent to Article 32 GDPR. Security duties arise from sectoral and state laws, such as the Health Insurance Portability and Accountability Act (HIPAA) Security Rule for electronic protected health information and the California Consumer Privacy Act, as amended by the California Privacy Rights Act (CCPA/CPRA), for covered California personal information. Requirements for image and video anonymization depend on the applicable state law, sector, contractual obligations, and the sensitivity of the footage.

Standards and references

Article 32 GDPR is the primary legal source for TOMs. The following publications provide useful frameworks for documenting and evaluating security measures for image and video processing:

See also