Core Definition
Video anonymization removes identifying elements (such as faces) from footage while preserving the signals analysis needs, like movement and pose. When irreversible, the anonymized data falls outside personal-data regimes — Korea PIPA (Article 58-2), GDPR (Recital 26), and the CCPA.
Full Anonymization vs. Partial Anonymization
Full anonymization erases the entire scene to maximize safety. Partial anonymization strongly erases only the identity-bearing regions, leaving analysis signals such as movement and pose intact. Which level to use depends on the purpose, not on one approach being uniquely correct.
The key is irreversibility. Simply blurring pixels is not enough. Data qualifies as "anonymous information" only when it retains the minimum values needed for analysis, discards the rest, and cannot be reconstructed back into the original.
The Difference That Order Makes
Even under the same label of "anonymization," a different order produces a different result. In a structure that analyzes the original footage first and deletes it later, the original footage and identity exist inside the system for the duration of the analysis. When anonymization is placed as the first step, the analysis system never holds identity at all.
This order is also a matter of trust. A structure that never holds the data in the first place is easier for both customers and regulators to verify than a promise to delete it later.
How SAAI Uses It
This is DEEPINGSOURCE's core technology. The SEAL engine anonymizes video in memory at the point of input, and the original footage is never stored anywhere in the system. Unlike vendors that analyze the original footage and delete it afterward, anonymization here comes before analysis. Every spatial analysis across the saai product line is built on top of that anonymized signal. This processing structure has been confirmed through an official interpretation from the Personal Information Protection Commission (August 2021) and outside legal review (December 2021).
See it in action