Over 170,000 people are part of the Sophos community on Facebook. Why not join us on Facebook to find out about the latest security threats.
Hi fellow Twitter user! Follow our team of security experts on Twitter for the latest news about internet security threats.
Already using Google+? Find us on Google+ for the latest security news.
Friend and former colleague Graham Cluley just drew my attention to an astonishing security problem with Xerox scanners.
I bet you it isn't the sort of problem you're thinking of, either.
These were probably things such as incompletely deleted data left behind on the hard disks of decommissioned scanners; poor security configuration in network-enabled scanners, such as default passwords; and exploitable vulnerabilities in image-handling code built in to scanners.
This problem is quite different.
(For all we know, this flaw may be present in other vendors' scanners, too, as it is a consequence of an algorithm chosen for compression. Xerox comes into it simply because that is the brand of scanner in the story.)
A Germam computer scientist, David Kriesel, was perusing the rooms depicted on some building plans he had scanned on a Xerox WorkCenter scanner.
He spotted an alarming anomaly: high quality errors.
That sounds odd, and it was.
Normally, when you notice scanning errors, it's because the quality is poor and the details illegible.
A room that is 15m2 on the original might looks like 1?m2 on the scanned copy, with the 5 scanned so badly it doesn't even look like a digit.
Or the 15 might be blurred, or have sufficiently many stray pixels in it, to look like an indecisive 16.
What you don't expect is that a crisply printed 21m2 on the original would be rendered as a crisply scanned 14m2, say, on the copy.
In other words, given the analog-to-digital nature of the scanning process, you'd expect imperfections, but you'd also expect the unreliable parts to look unreliable, thus making their unreliability self-documenting.
It turns out that the Xerox scanner in question was using a compression scheme called JBIG2, which emerged from the grandly-named Joint Bi-level Image Experts Group.
Bi-level images, as the name suggests, have just one bit per pixel, such as the images used in fax machines (if you remember them).
And JBIG2 has a clever, yet, with hindsight very reckless, feature: if two "swatches" of the image look like each other, the same data is used for both swatches, so that they effectively become identical.
This technique works perfectly in lossless compression, e.g. the deflate algorithm used in ZIP files, where the repeat of a string of characters such as NOTEWORTHY would be encoded as "repeat the ten characters I saw 164 bytes ago", not as another NOTEWORTHY.
But if imperfect matches were allowed, you might find NOTEWORTHY encoded as a repeat of NOT WORTHY, introducing an error that would be very hard to spot, despite the fact that the two phrases are antonyms.
The "fix", for our German computer scientist, seems to have been to use TIFF compression instead, a lossless image compression option supported by the scanner he was using.
Update: Xerox emailed us at 2013-08-10T11:15Z to point us at some official advice on the issue. In summary: JBIG2 compression isn't on by default. If you're worried someone might have changed the compression settings, a reset to factory defaults will change them back. Also, Xerox will be producing an optional patch that will prevent JBIG2 being turned on at all. (If you aren't sending faxes, you probably don't need it.)
The lesson to be learned here, other than that Graham has an excellent eye for interestingly quirky stories, is that algorithm choices are really important.
Instead of a blurry and obviously inconclusive image of the perpetrator, which would make it obvious that evidence would have to be sought elsewhere, you might end up with a clear and convincing image of someone who just happened to look like the perpetrator.
Where security is concerned, it's not just how safely you store what you've collected, it's how reliably you collect it in the first place.
Follow @duckblog
Earlier this week the World Wide Web Consortium (W3C) released the first drafts of two new privacy standards aimed at simplifying and standardising how websites read and comply with web users’ privacy settings.
Back in June the European Commission told the technology industry in no uncertain terms that if it didn’t agree a Do Not Track standard by the midde of 2012, it would be forced to act.
Since tracking tends to rely on cookies you might think that deleting them would be enough to prevent it. Sadly it's not. Researchers have shown that advertisers can work around users who delete their cookies by using Flash cookies or HTML 5 storage for the same purpose. Most creepily they can even resurrect delete cookies zombie-like using ETags.