Stripping EXIF From Images as a Security Measure

#fasterapp #infosec And you thought FourSquare was a security risk…

Mobile phones with great cameras are an awesome tool. Many of these end up on Facebook, visible to friends, family and, well, friends of friends and maybe even the public. They get shared around so much, you can’t really be sure where they might eventually wind up.

According to Justin Mitchell, an engineer for Facebook Photos, answering a Quora question on the subject last year, Facebook has “over 200 million photos uploaded per day, or around 6 billion per month. There are currently almost 90 billion photos total on Facebook. This means we are, by far, the largest photos site on the Internet.”

As most of these are uploaded via modern cameras – whether on mobile phones or digital cameras – which are almost universally enabled with GPS technology, they almost all certainly include some data that you might not want others to find: the exact location the picture was taken.

Pshaw! Many may think. After all, “checking in” via FourSquare and adding location to Facebook and Twitter posts is something many do regularly. But this data can be very dangerous, and not just for soliders who have been warned against geotagging photos uploaded to facebook, as cited by a recent Gizmodo article, “US Soldiers Are Giving Away Their Positions with Geotagged Photos”:

The Army has issued a warning to its soldiers to stop geotagging their photos on Facebook and other social media outlets. Because it's putting soldiers in danger, and has been for years.

Now you might not be worried about giving away the location of helicopters inside a compound that leads to the enemy able to “conduct a mortar attack, destroying four of the AH-64 Apaches” there, but the risks to everyone exists. Those who share photos of their home or things in their home (can’t resist showing off your latest collectible addition to friends, can you?) are opening themselves up to theft, especially if they also like to broadcast their latest travel schedules via a host of other socially connected tools. Even if you aren’t actively sharing your address, all a potential thief needs to do is grab a photo of your fat loot and extract the GPS coordinates hidden in the EXIF data to find his target and then move in, right after you made sure everyone know you were out of town by broadcasting your latest flight information (ATL –> ORD –> SEA).

“But I’ve locked down my photos using Facebook’s privacy features!” you say. You might have done so, but do you really know everyone on your list of “friends”? Are they really who they claim to be? And did any of them share your photo with their friends, and their friends? Facebook privacy doesn’t prevent the old standby of “save as” and “upload”, and a quick tag of your name and a Twitter search and bam! You’ve shared data with people you wouldn’t have, if only you had known.

While perhaps requiring a bit more paranoia than the average user (and an inherent distrust in humanity), there are very real security implications for a wide variety of folks to embedding Geotags in photos via EXIF, though perhaps those in service to their country more than others.

There is a simple and more automated mitigation for this risk. In addition to turning off Geolocation tags on your camera or phone or manually eradicating the EXIF info from photos, a mediating application delivery service can, on-demand, strip this data from images.

MEDIATED EXIF STRIPPING

With the right application delivery tier implementation, mediated EXIF stripping is as simple as other content scrubbing exercises. Requests are received as normal for an image object. When the image is retrieved from the origin server, a service in the application delivery tier is invoked that strips EXIF data from the image before it is returned to the end-user or deposited in a caching solution. Subsequent requests for that same image, then, though served out of cache are also clear of potentially dangerous GPS information – without modifying the original.*

That’s important, as for some folks having that information available to them may be necessary or desirable, but serving it up to the public may simply incur too much risk.

Given the velocity with which we click and share photos today, we may be underestimating the associated risk. Others may think that’s just far too paranoid and desire to keep EXIF data in their images. This is another opportunity to monetize a service for providers. The right application delivery tier, capable of interpreting context as well as being instructed by external infrastructure (including applications), could be configured such that only image-containing responses with specific HTTP headers are subject to EXIF stripping. The more security-minded users may desire such a service – and be willing to pay for it – while others could simply continue on as they were, EXIF and all.

And even if you aren’t concerned with potential security risks associated with EXIF, you might want to consider that stripping out that extraneous data from images like thumbnails and product shots can reduce the overall size of the image, which is a boon if you’re trying to improve overall performance – particularly on network and resource constrained devices like mobile phones.

* Image optimization techniques are always best-effort and sometimes cannot be applied to an image given other factors. Also, if a positive caching models is used, the original image is served the first time it is requested, but not cached.