Hacker News Reader

Comment by thegeomaster | Hacker News Reader

It seems like there needs to be another approach to redacting text like this. This one doesn't really cut it, apparently. And slips such as 'her' and 'she' was what primarily gave her away here. So the people who published the file should have been way more careful with this.

View on HN

Flash forward to modern times...

"It's okay, we're only collecting metadata."

Yes, which is to not redact text. As you have already found, it is an exceedingly complex problem to do it correctly. But more importantly, when it is done incorrectly, or redactions are made from political motivations (as we've seen excessive evidence of), they are practically impossible to challenge.

The difference of length between "he" and "she" would be enough to reveal that information.

They just draw a black box over the text in the PDFs now...

(No joke)

Yes, but now they remove the underlying text too. It was fun while it lasted.

Current best practice is actually to draw the black boxes, print it, then scan it back in. The result is crummy quality pdfs with no metadata or hidden text.

Well you'd think but I worked for an org back in 2005 that decided to completely optimise that step away.

They were an aerospace/defense company building IFF units no less...

I think that was still a year or so before news stories started coming out about government "redacted" documents that were not actually redacted.

A lot of the NSA-related documents released after redaction were certainly run through the "redact,print,scan" routine, and I know the Navy does the same as a best practice now.

Maybe that hasn't made it out to the defense industrial base by now, but I'd be surprised if 9 years worth of being beat about the head regarding redaction mistakes wouldn't have fixed things even there.

Wouldn't "redact, print-preview, save" be good enough?

Or redact, screen-shot.

Does that preserve anything that an actual paper print and rescan would remove?

Oh, there's tons of much better methods that would be more than good enough. Adobe even provides specific redaction features that drill down within the PDF data itself to make sure that nothing gets drawn under a redaction block, it works quite well.

But the government is about making procedures that are both easily understood by low-paid civil servants and unlikely to be screwed up in major fashion (e.g. our GS-7 scanning documents back in might use the wrong file name or reverse the pages or something, but they probably won't leak classified data if they do that).

Reply on news.ycombinator.com