Risk of Identity Inference From Backgrounds, Reflections, and Text in Images

Backgrounds, reflections, and text inside images can reveal routine places, identity, workplaces, schools, and people around you.

For image anonymity, looking only at EXIF and GPS is not enough.

Even if metadata is removed, identity and routine places can be inferred if information appears in the image itself. Background signs, the view outside a window, documents on a desk, uniforms, reflections, wall notices, and notifications on a screen. This information can be found through image search or by human eyes.

When protecting anonymity, you need to look at the "content" of the image.

This article explains how backgrounds, reflections, and text inside images become clues. A broader image check is covered in "How Images Can Break Anonymity."

Backgrounds Show Routine Places

Backgrounds contain a lot of information that shows location.

Station names, store names, road signs, school names, company logos, buildings, mountains or the sea, and distinctive interiors. Even if the person is not aware of them, someone looking may understand them.

Images can contain more information than text.

Before posting, enlarge the image and check the background.

Watch Reflections

Reflections are easy to overlook.

Mirrors, windows, eyeglasses, smartphone screens, metal, cars, glass, and glossy desks can reflect the photographer, the opposite side of the room, screens, and documents.

If a face appears in a reflection, hiding the face elsewhere is meaningless.

Check reflections both before and after image editing.

Text Information Becomes a Strong Clue

Text inside an image is a very strong clue.

Name tags, envelopes, documents, notices, receipts, slips, whiteboards, tab names, notifications, and filenames. Even if they are small, enlargement or OCR may make them readable.

"It is too small to read" is dangerous.

After posting, assume that other people will enlarge it, adjust the brightness, and use OCR.

Blurring Has Limits

Blurring part of an image through image editing can be effective.

However, it is dangerous when the blur is weak, the area is too narrow, the surrounding information still allows inference, or the original image remains somewhere else.

For high-risk images, it may be necessary to choose not to publish the image rather than partially blurring it.

Pre-Posting Check Flow

Before posting an image, turn the visual check into a procedure.

When you are used to the image, it is easy to stop after checking only the face or large text. But in anonymity, you also look at small text, reflections, the edges of the background, notifications, and filenames.

When using image search, sending the image to an external image search service is itself a risk. Do not upload unpublished high-risk images, reporting materials, internal materials, or images containing victims or minors to an external service as-is. Even when necessary, check with a copy edited down to the range that is safe to publish, or prioritize methods that do not send it externally.

Images may be cropped and reused after publication.

Even if the whole image looks safe, information may be readable when part of it is enlarged. Before posting, try cropping and enlarging it yourself.

Difference From Metadata Checks

This check is separate from checking EXIF and GPS.

Even if EXIF is removed, a station name in the background reveals the location. Even if GPS is removed, a store name and time on a receipt reveal activity.

For image anonymity, check both internal information and appearance.

Be Careful With Other People's Information

Images also show information about people other than you.

Passersby, family, colleagues, children, license plates, name tags, and documents on desks. If you edit only with your own anonymity in mind, you may involve people around you.

In anonymity, protect not only yourself, but also the people and related parties shown in the image.

Also check the purpose of publishing the image.

Is it necessary as evidence, do you only want to convey atmosphere, or can an explanation in text substitute for it? The weaker the purpose of an image, the harder it is to justify its anonymity risk.

Images pass more information at once than text.

Information that could be handled by deleting one word in text may appear in an image as background, reflection, text, time, and people all at once. In posts that prioritize anonymity, always consider "can this be replaced with text?" before publishing an image.

When Showing Only Part of an Image

Even when you need to publish an image, you do not necessarily need to publish the whole image.

If you crop only a document section, a broken part, or the range needed to explain a situation, you can reduce background and reflections. However, surrounding text, shadows, reflections, screen aspect ratio, and filenames can remain even after cropping.

For image anonymity, deciding to "show only the necessary range" is more effective than editing technique.

Create a publication image separately from the original image, and recheck only the edited image.

Summary

For image anonymity, check not only EXIF and GPS, but also the background, reflections, and text inside the image.

Even if metadata is removed, routine places and identity can be inferred if signs, documents, notifications, the view outside windows, mirrors, eyeglasses, or filenames are visible.

Before posting, enlarge the image, check brightness and reflections, and see whether text can be read.

Blurring is effective, but it is not all-powerful.

In anonymity, before editing an image, first consider whether that image needs to be published at all.

Related tools

Related articles

• Metadata —How Images Can Break Anonymity
• Metadata —Removing image metadata
• Metadata —Metadata left in submissions and file sharing
• Metadata —Checking SVG metadata
• Metadata —SVGO and manual XML checks
• Metadata —What Is Metadata?