Obfuscation vs redaction: what is the difference?

Here is the whole distinction in one line. Redaction removes information so it cannot be recovered. Obfuscation only obscures it, and obscured information can be brought back. That difference is the entire game when the thing you are hiding sits inside an image. A blur, a pixelation, a low-resolution mosaic or a semi-transparent box all obfuscate: they leave the original pixels in the file, changed but recoverable. A solid opaque fill flattened into a new image, or a content-independent scramble that holds no relationship to what sat underneath, redacts: the pixels are actually gone. If you remember one thing, remember that a screenshot you obfuscated is a screenshot someone can still read.

The definitions, kept straight

The two words get used as if they mean the same thing, and in casual use people call any dark box over a secret a redaction. They are not the same, and the gap between them is exactly where secrets leak. Redaction is the removal of information from a file so that it cannot be recovered from that file. Obfuscation is any change that makes information harder to read while leaving it present in some recoverable form. Data-security writing draws the same line for databases and documents, where obfuscation covers reversible techniques like tokenization and encryption and irreversible-but-lossy ones like masking. For an image the picture is simpler and more brutal: either the pixels are gone, or they are still there under a disguise.

Why the distinction is the whole game for a screenshot

With text you can delete a word and there is nothing left to find. An image is different. A screenshot is a grid of pixels, and most of the things people reach for to hide part of it do not delete those pixels, they only alter them. A blur runs a math filter over the region. A pixelation averages each block. A translucent box tints what is beneath it. In every one of those cases the original signal is still in the file, just scrambled by a process that can be studied and undone. That is why the same black rectangle can be a real redaction or a useless disguise depending on one invisible detail: whether it was flattened into the pixels, or is sitting on a layer over intact data. If you want the step-by-step safe method, I wrote how to redact a screenshot on a Mac. This page is about why it matters.

Obfuscation only obscures, and obscured pixels come back

Every common way to hide part of a screenshot is obfuscation, and each one has a documented way back. A blur can be estimated back, because a low-radius blur is a reversible filter, so it obscures but never removes. Pixelation is worse, not better: each block is the average of the pixels beneath it, a fixed mapping a solver can walk through. Depix showed this in 2020, and in 2022 Dan Petro at Bishop Fox released Unredacter, which reconstructs pixelated text by rendering candidate words in the same font, re-pixelating each guess, and matching it against the image character by character. It works best on short structured strings in a known font, which is exactly the account numbers and codes people pixelate. Petro's advice was blunt: no pixelization, no blurring. A semi-transparent box leaves the text readable once someone pushes brightness and contrast, and a box on its own layer can simply be peeled off or the text copied out from underneath. I go deeper into what can and cannot be undone in can you reverse a blurred or pixelated image.

Obfuscation vs redaction, method by method

Here is the same idea as a table. The only column that matters in practice is the last question: can the original be recovered?

Redaction removes the pixels for good

Only two things on that list actually redact. The first is a solid, fully opaque fill flattened into a fresh file. Once you export a new image, the cover is baked into the pixels and nothing sits on a removable layer, so there is no original left underneath to recover. The second is a content-independent scramble, which I explain below. Both share the property that defines redaction: the information is gone from the file, not disguised inside it. The NSA's guide, "Redacting with Confidence," makes the same point in plain terms. Its central idea is that information hidden or covered in a computer document can almost always be recovered, so the fix is to remove it, not obscure it. That guidance was written for Word and PDF, but the principle is identical for a PNG.

Is obfuscation ever the right tool?

Yes, when the goal is not to hide one specific secret forever. Obfuscation earns its place where you want to make data hard to use while keeping it workable: obfuscating source code so it is painful to reverse-engineer, or masking real values in a test database so developers can work without touching live records. Those are legitimate uses, and some of them are deliberately reversible with a key. The mistake is borrowing that toolkit for a screenshot. If the value in the frame is a password, an API key or a client's account number, obscuring it is not enough, because someone with the recovered file gets the value back. That is also why "can redaction be reversed" is a trick question. Real redaction cannot, but most of what people call redaction is obfuscation wearing the name, and that is exactly what the aCropalypse bug exposed when cropped-out image data survived inside files people believed were clean.

How ScrubShot draws the line

I built ScrubShot around this exact distinction, because the safe path is fiddly to do by hand and easy to get half right. Its Scrub tool is real redaction, not obfuscation. A normal pixelate sets each block to the average of the pixels beneath it, which is the signal Depix and Unredacter exploit. ScrubShot does not do that. It uses a content-independent scramble: each mosaic block is filled from pixels sampled at random from across the whole selected region, so the block pattern is decoupled from whatever sat underneath. It looks like a redaction and keeps the rough colors of the area, but there is no per-block relationship to the original to solve for. It reverses to nothing by design, and the scramble is baked straight into the image, so the original pixels are gone rather than hidden on a layer.

The whole capture, scrub and share flow makes no network connections, so a screenshot with a secret in it never leaves your Mac. To be precise, the app talks to a server only for the free-trial check, license activation and update checks, and that request carries a hashed machine identifier and your license key, never any image data. ScrubShot runs on macOS 14 and later, on both Apple Silicon and Intel.

FAQ

What is the difference between data redaction and obfuscation?

Redaction removes the information so it cannot be recovered. Obfuscation only obscures it, and obscured information can be brought back. In an image, a blur, a pixelation or a semi-transparent box is obfuscation, so it is recoverable. A solid opaque fill flattened into a new file, or a content-independent scramble, is redaction, so the pixels are actually gone.

What are examples of obfuscation?

In an image, obfuscation is a blur, a pixelation or mosaic, a low-resolution downscale, or a semi-transparent box laid over the secret. Each one only obscures the pixels. A blur can be estimated back, pixelation can be reconstructed by tools like Depix and Unredacter, and a box on its own layer can be peeled off.

Can redacting be reversed?

Proper redaction cannot be reversed, because the information is gone from the file. The problem is that a lot of what people call redaction is really obfuscation, and that is reversible. A blurred or pixelated area, or a black box that was never flattened, can often be recovered, which is how bugs like aCropalypse exposed content people thought was redacted.

Can obfuscation be used legally?

Obfuscation is a valid tool where the goal is to make data hard to use rather than to hide one specific secret forever, such as code obfuscation or masking a value in a test database. It is the wrong tool for a screenshot secret. If someone must never read the value, you need redaction that removes it, not obfuscation that only obscures it.