Is it possible for someone to see under the "blacked out" part of this image (see below)?

Question

Ok, so I sent some pictures to a journalist to report something I thought was noteworthy. However, one of the images contained my date of birth and other personal information so I blacked it out using markup tools in the Photos app of my iPhone 6 (running iOS 10.3.2). Then, I took a screenshot, then sent the screenshot from my phone to my email account, then downloaded the image to my laptop. Then, I sent that image from my desktop to the journalist.

So, let's say this was the edited original picture (it's in JPEG format because when I sent it from my iPhone via iCloud mail to my other email address then downloaded it on to my laptop from that other email address, it downloaded in JPEG, but the screenshot downloaded as a PNG):

https://drive.google.com/file/d/0BylJxh-1HrrkQUFqWXprbHhNVUU/view?usp=sharing

This is the screenshot (PNG format):

https://drive.google.com/file/d/0BylJxh-1HrrkREZWRFdHOUxnTTA/view?usp=sharing

Using whatever technique you want, is there any way possible way to see what is under the blacked out parts in the screenshot, because that was what I actually then sent to the journalist.

Thank you for your help!

PS - The images above are examples and not the actual thing I sent! It just says "LALALALALA BATMAN!"

Answer 1

PNG is a bitmap format, thus "blackening out" is a destructive method: what was in the original picture before it was replaced by black pixels is no more in the blackened picture. You can use any fully opaque color (a color without a transparent value) for this process, not just black.

This is not the same for other formats that are layered, like PDF, SVG (vector image), PSD (photoshop image), etc. In those formats, "blackening out" will most likely add a black layer that can easily be removed later.

PNG doesn't have EXIF metadata like JPEG, but still can embed some king of metadata in the ancillary chunks. You would be wise to check what's contained in those chunks of your picture.

Answer 2

There are ways, but possibly not in your example case (blackeneing at end of line).

If you blacken a word or short phrase (e.g., a name) in the middle of a paragraph of justified text in a proportional font then an attacker may be able to undo this. The reason is that the spacing between the other words in the same line may give away how many pixels exactly the hidden word takes. And for some fonts this length may be different for virtually all plausible candidate words.

Answer 3

Assuming the virtual paint you used to black out the words in question is 100% opaque, and you painted over every single pixel you want to hide, and you sent a screenshot of the result, there is no way to restore the original pixels.

Any method to recover text under the blacked-out parts would need to rely on surrounding context, as Hagen and Overmind point out.

Answer 4

It's not safe to use Markup's pencil or marker tools for redaction because they are not opaque. (Even if you use the opacity slider at its max setting.)

If you search "ios markup redaction" you'll probably see PSA: iOS Markup is not designed to be a redaction tool for sensitive information. That article recommends against using the pen tool, even though they say it is opaque, because it's hard to completely write over text with such small lines. However I wouldn't be shocked if the pen tool was semi-transparent in other iOS versions.

I suggest importing the image in GIMP, manipulating it, then export in JPEG or PNG format. The fool proof method would be to use the rectangle selection box, press delete, and fill (without deselecting) the same space black.

Flood filling the space in with black doesn't redact anything more than what the deletion deletion did. Changing the color to black is just for the recipient's benefit. (Make the redactions more obvious.) I would not rely on the flood fill alone on the off chance that I accidentally use the wrong tool settings.

Importing the image into GIMP probably does the same thing as what you want taking a screenshot of the marked-up image to do. (Strip the original metadata and history from the photo.)

If you're paranoid you might worry about proprietary image editors embedding watermarks or steganographic information. (Or even if they're not paranoid you may worry about them unintentionally leaking some information in an exported image.) If you feel safe using MS Paint or Photoshop then you may use them instead. Or even Markup. The article I linked to says

It is possible to add a rectangle shape with a solid color. Tap the +, select Square, position, and then tap the options button and change the style from outline to a filled in shape. This works fine, and not susceptible to the marker transparency issues, although it is quite buried in the interface.

I would not trust a different smart phone app. The risks outweigh any convencience. even if it's not a huge problem if redacted data is leaked.

Remember to make the rectangle a few pixels larger than the text itself, especially if it's a JPEG.

Answer 5

You can determine this given a certain font is used and of course assuming that the black area did not use a different font/size. There are heavy limitations though. Some fonts use just a few constant character-occupied area while some adapt the occupied area for each character.

You can also determine a black word or group of words by analyzing the rest of the text. Won't work for specific dates or names, since you cannot extrapolate such things from the general text content (so it does not apply in your case).

Answer 6

As others have said, using semi transparent marker is probably the most common mistake that one could be making when redacting documents, but there are other more obscure avenues of attacks as well.

If you compressed the text using lossy image compressions like JPEG, then the image can have some compression artefacts, which in theory could leak some information about the underlying text. When redacting texts, especially on heavily compressed image, you may need to redact a slightly larger area than just the text, to avoid side channel leaks from these compression artefacts.

At the most basic level, with just the most basic of image viewer, compression artefacts attacks may allow one to rather easily discern whether the letters contains any ascenders or descenders, and if combined with other contextual clues, it may allow the attached to guess the blacked out texts. Discerning any more text beyond that should be possible in theory, though it'll require much more complex math and specialised tool to enumerate the possible letters based on the compression artefacts that those letters create.