Third-party data - definition
Third-party data is information about a person other than the individual making a request, providing material, or otherwise acting as the primary subject of a file. In image and video anonymization, it commonly means information about other identifiable people appearing in the same photograph or footage.
For example, a person may request a copy of security camera footage showing an incident involving them. The footage can also show bystanders, employees, visitors, vehicle occupants, faces in the background, and readable license plates. Information that identifies those other people is third-party data in relation to the requester. Before the footage is disclosed, shared, published, or used for a new purpose, the organization must assess whether those third parties need to be redacted.
Third-party data is an operational privacy term, not a single statutory category with one universal definition. Its treatment depends on the purpose of processing, the recipient, applicable privacy law, public-records rules, contractual obligations, and the risk that a person can be identified from the image or video.
Third-party data in images and video footage
In visual material, third-party data can be visible directly or inferred from the context of the recording. A redaction review must therefore consider both obvious identifiers and details that can identify a person when combined with other available information.
- Faces: A face can identify a bystander, customer, employee, patient, student, or other individual appearing in the scene.
- License plates: A license plate can be linked to a vehicle owner or user through records or other information available to a recipient.
- Contextual identifiers: A distinctive location, vehicle, clothing, role, timestamp, or sequence of events may make a person identifiable even if their face is partly obscured.
- Visible sensitive content: Documents, name badges, computer monitors, tattoos, and company logos may require review where they reveal personal or confidential information.
Face detection and license plate detection are useful first steps, but they do not decide whether disclosure is lawful. They identify visual regions that may require masking. A qualified reviewer must assess the footage, the intended recipient, and any applicable disclosure exception.
When third-party data requires redaction
Redaction is generally required when a disclosure would reveal identifiable information about a person who is not entitled to receive it, and no applicable legal basis, consent, exemption, or disclosure obligation permits the release. The required scope is not always the same as the scope of automated detection.
A practical review can use the following sequence:
- Identify the requester, recipient, and permitted purpose of the disclosure.
- Determine whether the image or video contains people other than the requester.
- Locate direct identifiers, including faces and license plates.
- Review contextual details that could identify a third party after release.
- Apply irreversible masking, cropping, replacement, or another appropriate redaction method.
- Perform quality assurance on the exported file, including adjacent video frames and scene changes.
- Document the decision criteria without retaining unnecessary personal data in processing logs.
For video, a redaction must persist for every frame in which the person or vehicle is visible. A mask that misses even a small number of frames can expose the third party. This makes temporal tracking, scene-cut handling, and final visual review important parts of the process.
Detection quality and redaction quality
Detection performance and redaction performance are related but different. A model can detect most faces while still producing an inadequate disclosure copy if masks are misplaced, too small, or absent after an occlusion or scene change.
Measure | Meaning for third-party data | Practical implication |
|---|---|---|
Recall | The proportion of relevant faces or plates detected by the system. | Low recall creates a risk that third-party data remains visible. |
Precision | The proportion of detected regions that are actually faces or plates. | Low precision increases unnecessary masking and review work. |
Mask coverage | Whether the redaction fully covers the identifying visual region. | Partial coverage may leave distinctive features visible. |
Frame continuity | Whether masking remains effective across consecutive frames. | Failures can expose a person during movement, occlusion, or camera motion. |
Recall can be expressed as TP / (TP + FN), where TP is the number of correctly detected target regions and FN is the number of target regions missed by the system. For disclosure workflows, missed detections usually present a higher privacy risk than false detections. The National Institute of Standards and Technology (NIST) describes precision and recall as standard evaluation measures for information retrieval and detection tasks.
Automation and manual review
Automated face blurring and license plate blurring can reduce the volume of manual work in large image and video collections. However, automation should be configured as part of a controlled redaction workflow rather than treated as a legal determination.
Gallio PRO automatically detects and blurs faces and license plates in supported image and video processing workflows. It does not automatically detect company logos, tattoos, name badges, documents, or content displayed on monitors. These elements can be redacted manually with the built-in editor when the disclosure assessment identifies them as third-party data or confidential content. Gallio PRO does not perform real-time anonymization or video stream anonymization.
Standards and references
The following sources support a risk-based approach to identifying and protecting third-party data in visual material. They do not replace a legal assessment of a specific disclosure request.
- NIST Privacy Framework 1.0, National Institute of Standards and Technology, 2020.
- NIST Special Publication 800-122, Guide to Protecting the Confidentiality of Personally Identifiable Information (PII), National Institute of Standards and Technology, 2010.
- ISO/IEC 20889:2018, Privacy enhancing data de-identification terminology and classification of techniques, International Organization for Standardization and International Electrotechnical Commission, 2018.
- California Consumer Privacy Act (CCPA), California Civil Code § 1798.100 et seq., as amended by the California Privacy Rights Act.