When you "edit" a raw image with pretty much all available raw processing applications, you're not really editing the raw data. You're editing the set of instructions on how that data should be used to produce a viewable image. The actual raw data is not altered.
When you look at a 'raw' image on a screen, you're not really looking at "THE raw image", either. You're looking at one possible interpretation among a (practically infinite) number of equally valid interpretations of that data.
If someone else opens a raw image file that you have "edited" to blur or black out faces, your edit will only be used and the faces blurred or blacked out if the application the other party uses to open the raw file:
- Understands the instructions you added with your previous edit. This usually means the same application is used to by both you and the other party
- The application on the second user's device is set to use those instructions, rather than some other set - such as another default set - to open the file.
Even then, the user could easily reverse your blurring edit to see the faces without the blur.
If the raw file is opened by a different application than the one you used to blur or black out the faces, it is nearly impossible that the faces will be blurred or blacked out in the image displayed on the screen.
I was wondering if there was a way to actually obscure or delete the information in some pixels in the raw data.
If the original raw data is unaltered by every commercially available raw processing application and by every popular open source raw editing/viewing application, and you are using any such commercially available or open source application, the answer should be fairly obvious: "No."
I suppose one could create an application that could decompress the raw image data to a raster format, convert the monochrome luminance values of each pixel in the raw file to color, apply gamma correction (so that you can see what you are doing instead of trying to work on a black blob of nothingness), mask out the faces, and then convert the gamma corrected values back to linear values and re-encode that into whatever particular raw file format your images use (NEF, CR2, etc.).
Or, as others have pointed out, one could designate locations to be altered based on the preview jpeg image usually provided in the raw file.
One would still need to decompress the raw image data into a raster image, apply the changes to the specified locations, and then compress the raster image back into the original raw file format using whatever specific compression scheme the original raw format uses. One would also need to delete the original jpeg preview (and any other preview/thumbnail jpeg or tiff images appended to the original raw file) and replace it (them) with an altered one, either generated from the altered raw data or by saving the changes made to the original jpeg preview.
But to the best of my knowledge, no such product is commercially available. I'm also not aware of any open source product that can do that.