Hacker News Reader

Comment by toothbrush | Hacker News Reader

This sounds like a job for an approach like [0]! Reminds me of [1].

0. http://lcamtuf.blogspot.fr/2014/11/pulling-jpegs-out-of-thin..., http://lcamtuf.coredump.cx/afl/, american fuzzy lop

1. http://news.cornell.edu/stories/2015/03/images-fool-computer..., Images that fool computer vision

Doesn't AFL rely on being able to compile the application with it's hooks? It's been a long time since I've messed with Photoshop, but it's a complete mess of random dead code inside (even includes poems about a dead dog believe it or not), chances are you're looking for a completely obfuscated bunch of assembly in the hundred plus megabyte binary. I love Hopper, but not that much.

Yes, you're right. I'm admittedly not at all an expert on this type of thing, but surely some randomised testing, starting from a real banknote and making small mutations, would be doable.

Oh sure, you could make a quick imagemagick script to make random variations and get a pass/fail with AppleScript, or even do some exploring with a debugger while opening one. Main barrier is motivation, I have no desire to counterfeit money (and who would use PhotoShop for that anyway), so I decided to let that rest.

It wouldn't be about counterfeiting, but about learning more about the detection algo :)

I'm mostly concerned about experimenting straying into "reverse engineering".

Indeed, the FBI are known to lack a sense of humor when it comes to funny money.

The secret service, even less so.

Just make sure you counterfeit so many bills they offer you a tiny punishment in exchange for helping them track them down

http://www.dailymail.co.uk/news/article-2624697/Master-count...

According to wikipedia [0], the code that recognizes money was given to Adobe as a binary blob. So no go even if you have acces to the Photoshop source code. [0] http://en.m.wikipedia.org/wiki/EURion_constellation

You could run PS within QEMU or something, perhaps. Would be slow though.

Reply on news.ycombinator.com