Anonymizer Module

Erase the identity.
Keep the analysis. Anonymized on input.

Identity disappears as video enters. That is what makes movement, posture, and dwell safe to analyze for operational decisions.

TechnologyAnonymizer

Mechanism

Four steps that erase identity

Not who, but what and how. Anonymizer never blurs the whole frame. It erases only the regions that reveal identity, on input, and leaves every signal analysis needs intact.

01Input
02Detect identity
03Anonymize on input
04Forward signal

Before-and-after comparison of the same person, anonymized — Drag the handle — anonymized at the first step. The original is not retained.

Drag the handle — anonymized at the first step. The original is not retained.

Find the identity

In every frame, it locates the regions that could single out a person — faces, body shapes, license plates. Not to learn who they are, but to mark what must be erased.

Anonymize on input

Only the identity regions are erased, irreversibly, while the signals analysis lives on — movement, posture, dwell — are preserved. Region by region, so what happened and how stays fully readable.

On-device processing

Computation can finish on edge devices or locally. Raw frames never leave the premises; only the already-anonymized result moves to the next step.

Forward the signal

The anonymized stream and its analytical signals pass to the spatial-intelligence and agent stages. Analysis runs all the way through — the original footage is stored nowhere.

We read information, not identity

The face is erased; the context is read.

Identity is erased beyond recovery; only the minimum features the AI needs to understand the scene are kept. So even on anonymized footage we read gender and age range, movement and posture, attention and dwell — without ever knowing who.

The face is noise. Still readable — age range, gaze, speaking. No identity.

SEAL · Privacy by design

We see what is happening, not who is there.

No original kept

Anonymized at the first step. The original is left nowhere in the system.

No human viewing

No human eye reaches it. The machine sees.

No re-identification

Identity cannot be recovered from anonymized data.

Across Environments

Any space — only identity is erased

It works the same way across very different environments — factories, hospitals, cinemas, schools. People are anonymized while the space and the situation stay readable. Drag the handle to compare with the original.

Factory-floor footage before and after anonymization — Factory

Hospital footage before and after anonymization — Hospital

Cinema footage before and after anonymization — Cinema

School footage before and after anonymization — School

Why not the usual way

The dilemma of conventional de-identification

Masking and blur destroy the data's value; face-swapping is hard to verify; using the original leaves risk behind. SEAL avoids all three — identification made impossible, analysis kept at original-grade.

	Legal compliance	Data utility	De-ID accuracy
Use original	Falls short	Meets	Falls short
Masking & blur	Partial	Falls short	Partial
Face swap	Partial	Partial	Falls short
SEAL	Meets	Meets	Meets

Specification

Specification overview

Item	Detail	Note
Input	RTSP / file / frame stream	On the CCTV you already have
Processing unit	Identity regions within a frame	What and how stay intact
Preserved signal	Movement, posture, dwell, density	Analysis runs all the way
Erased	Identity such as faces, body shapes, plates	Who is never kept
Deployment	Edge / on-premises / server	Chosen to fit the environment
Original footage	Never stored	Anonymized result only
Integration	Spatial-intelligence and agent stages	First step of the operating loop

The values above describe the configurable range; actual specifications vary with the deployment environment.

FAQ

Frequently asked questions

No. Identity is erased the moment footage comes in, and from there only the anonymized result stream and analytical signals move downstream. Raw frames never remain in storage — that is the default.

Re-identification is impossible by design. Identity regions are erased irreversibly while analytical signal is preserved, so who someone was cannot be recovered from anonymized data. Processing strength is tuned to your environment.

Yes. Everyone in frame is anonymized at once, regardless of consent — including the hundreds who pass by a storefront. No one is left identifiable; only the flow is analyzed.

It accepts standard video input such as RTSP, so it runs on the CCTV you already have — no camera replacement required. The specific compatibility scope is advised after an environment review.

Erasing identity on input — privacy by design — aligns with the data-minimization principles GDPR, CCPA, and Korea’s PIPA require, and we hold SOC 2 certification. Final compliance depends on your overall data-processing scheme, so we recommend reviewing it together at adoption.

It is designed for real-time stream processing; the level of latency varies with input resolution, frame rate, and deployment hardware. Expected latency for your operating environment is estimated together during the specification review.

See all FAQs

Anonymization review

Let’s design an anonymization setup for your space

Compatibility with the CCTV you already have, deployment form, regulatory requirements — answered directly by our technical team.

Request a technical review See SEAL SDK

Erase the identity. Keep the analysis. Anonymized on input.