Geo guessing is related to open source intelligence techniques, and it’s pretty easy to get surprisingly good at it.
People who are good at it can take a picture of someone’s room and deduce enough about them (sometimes) to be able to get their name, address and phone number.
It being automatic is pretty cool, but you were already leaking the information to anyone interested.
This Just In: Most photos uploaded to the internet are not stripped of their metadata, and one of the common things kept in metadata is… (drumroll please)… your GPS coordinates.
This is a lot less interesting than it seems to be at first glance, imho.
Yeah, I have it for personal photos that will never be shared. If I am traveling, I want a record of where a given photo was. But those aren’t photos I am sharing, and the ones I do share get their metadata stripped.
Some digital cameras and phone cameras can also embed the GPS coordinates in the pixel data so that even if you delete the EXIF metadata the GPS location and device serial number are still present in the image. Many document printers also embed device serial number and other data on printed documents by using nearly invisible dot encodings.
No easy way at all. The specs would be in-house manufacturer docs. Recall that digital cameras used to embed date and time visibly in images in a corner. The logical progression was to embed other data such as device serial number, geotag data, etc.
Regarding the schemes for steganographic identification in devices such as cameras and printers, this information is usually kept a trade secret. The Secret Service would probably already have the spec docs for data hiding. Many manufacturers already have working agreements to provide back door assistance and documentation for the hardware surveillance economy. Ink chemistry profiles are registered with the Secret Service. The subterfuge is to 'investigate counterfeiting' but it is also used to identify whistleblowers and objective targets by their printer serial number or ink chemistry, or the data embedded in any images they are naive enough to publish.
If you are a undercover reporter secretly video recording, unbeknownst to you the video could have metadata encoded using a secret scheme. If you registered that product for a warranty, or bought it online and had it shipped, or paid with a credit card or check, or walked beneath the electronics store cameras without a hat and sunglasses to pay cash, it is easy for the state organs to then follow the breadcrumbs and identify the videographer.
Almost all 'free' wifi hotspots offered by chain restaurants and hotels are logged with the data being stored indefinitely, showing your mac address. It takes only a little bit of investigation and process of elimination to find the user on a camera feed history, to see who was connected when a certain message or leak was sent. If you use a wifi hotspot in a McDonalds, Wendy's, Starbucks, etc. smile for the surveillance camera which will also have your device's unique MAC address in the wifi history. This MAC address data is automatically sent to a central station, for example at the Wandering Wifi company, and God only knows how long they store it.
None of this nonsense makes anyone safer. These people hate us.
I think we can trust that most phone camera apps do in fact obey the toggle they provide for whether or not to embed the GPS location data in the image.
Back in like 2006 or 7 steganography was used in obscure corners of the internet ( like insurgen.cc, an early anonymous holdout that got broken up by the feds) to pass around hacking tools. You’d unzip the dangerous kitten photo with winrar and extract a set of hacking tools. One I remember passed around widely was the low orbiting ion cannon the /b used to ddos scientologists.
I think I have read that on some versions it can store the app’s package name in the metadata. Not sure if that counts private but if and when it does so, it’s good to be aware of
Yes, this is a privacy thing, we strip the EXIF data. As long as you’re not also adding location to your Tweet (which is optional) then there’s no location data associated with the Tweet or the media.
People replying to a Twitter thread with photos are automatically having the location data stripped.
God, I can’t wait for LLMs to automate calling out well intentioned total BS in every single comment on social media eventually. It’s increasing at a worrying pace.
I’m sure most people who would put this to test would strip that data or screen grab the image to do the same thing…. If you know about meta data, so does a large amount of other people mate…
The people would be labeled as a fraud very fast if this wasn’t actually a real thing dude.
It’s just sourcing data from Street View or similar. Not that scary. If it picked you out of a crowd in a randomly sourced image from that area, then it’d be scary.
It really isn’t that hard if anything like a silhouette of mountains are in the background and you have a couple of rough hints that give you an idea where to start or how to narrow down possible locations, no AI needed.
You’re misunderstanding the post. It’s not about whether or not someone could guess your location from a picture. It’s about the automation thereof. As soon as that is possible it becomes another viable vector to compromise your privacy.
True, but that just turns into a cat an mouse game. Also, one the photo is up, the background doesn’t change how its blurred with time --> wait long enough and a technique to unblur will be developed.
I do remember 1-2 years ago there is a paper (or model?) that reverse blured images. It’s similar to how ML based object remover and inpainting works. Granted it only works for specific blurring algo.
Some blues are reversible, and some aren’t. Some of them do a statistical rearrangement of the data in the area being blurred that’s effectively reversible.
Think shredding a document. It’s a pain and it might take a minute, but it’s feasible to get the original document back, give or take some overlapping edges and tape.
Other blurs combine, distort, and alter the image contents such that there’s nothing there to recombine to get the original.
A motion blur or the typical “fuzzy” blur can be directly reversed for the former, and statistical techniques and AI tools can be used on the later to reconstruct, because the original data is still there, or there enough that you can make guesses based on what’s there and context.
Pixelating the area does a better job because it actually deletes information as opposed to just smearing it around, but tools can still pick out lines and shapes well enough to make informed guesses.
Some blurs however create a random noise over the area being blurred, which is then tweaked to fit the context of whatever was being blurred.
Something like that is impossible to reverse because the information simply is not there.
It’s like using generative AI to “recover” data cropped from an image. At that point it’s no longer recovery, but creation of possible data that would fit there.
The tools aren’t magical, they’re still ultimately bound by the rules of information storage.
Yeah, for the pic you used as example, the tool will just create something that fits. Not really “unblur” the image but guess what it would be with the info it have. It will be very likely not the same face versus the original.
However, recreating background maybe easier and accurate enough for a geo guesser or a ML model to figure out roughly where the image was taken.
You do realize that a lot of image recognition was done on scaled down images? Some techniques would even blur the images on purpose to reduce the chance of confusion. Hell, anti-aliasing makes text seem more readable by adding targeted blur.
Deblurring is guessing and if you have enough computing power with some brain power (or AI), you can reduce the number of required guesses by erasing improbable guesses.
And you misunderstand my point, it always has been a way to compromise your privacy. Privacy matters most in the individual case, with people who know you. If you e.g. share a picture taken at your home (outside or looking out of the window in the background) with a friend online you always had to assume that they could figure out where you lived from that if there were any of those kinds of features in there.
Sure, companies might be able to do it on a larger scale but honestly, AI is just too inefficient for that right now, as in the energy-cost required to apply it to every picture you share just in case your location might be useful isn’t worth it yet.
Privacy matters most in the individual case, with people who know you.
That statement is subjective at best. My friends and coworkers knowing where I live certainly isn’t my concern. In today’s day and age privacy enthusiasts are definitely more scared of corpos and governments.
isn’t worth it yet.
You’re thinking too small. Just in the context of the e2ee ban planned in europe, think what you could do. The new law is set to scan all your messages before/after sending for specific keywords. Imagine you get automatically flagged and now an AI is scanning all your pictures for locations and contacts and what not. Just the thought that might be technically possible is scary as hell.
Governments won’t scan all your pictures to figure out who you are, they are just going to ask (read: legally force) the website/hoster where you posted that picture for your IP address and/or payment info and then do the same with your ISP/payment provider to convert that into your RL info to figure out who you are.
And you might not be worried about your RL friends or coworkers but what about people you meet online? Everyone able to see your post on some social media site?
Nobody is going to scan all the pictures you post for some information that is going to be valid for a long time after it is discovered once. Governments and corporations have had the means to discover who you are once for a long time.
this is extremely scary if true. are these algorithms obtainable by every day people? do they work only in heavily photographed areas or do they infer based on things like climate, foliage, etc? I would love some documentation on these tools if anyone has any.
There are tons of machine learning algorithm libraries easily usable by any relatively amateur programmer. Aside from that all they would need is access to a sufficient quantity of geographically tagged photographs to train one with. You could probably scrape a decent corpus from google street view.
The obtainability of any given AI application is directly proportional to the availability of data sets that model the problem. The algorithms are all packed up into user friendly programs and apis that are mostly freely available.
It might be easier to train the AI to the specific things Geoguessr players have collected as signs that give away a location instead of letting the AI figure all those out again.
Basically a combination of what the game geoguesser does, and public geotagged images to be able to get a decent shot at approximate location for previously unseen areas.
It’s more ominous when automated, but with only a little practice it’s easy enough for a human to get significantly better.
EDIT: yup, looks like this is the guy from the Twitter: andrewgao.dev and he’s Stanford affiliated with the same department that made the above paper and system.
I am not sure it’s the same software, but it’s a fairly good guess I think. Same software capabilities and same lab, with the same area of research.
Geoguesser is a subset of the skills used for general image geo location for open source intelligence.
In the specific cases of only using the data present in the image and relying on geographic information, it certainly does better.
Humans still do better, and can reach decent skill with minimal training, at placing images that require spatial reasoning or referencing multiple data sources.
AI tools will likely be able to learn those extra skills, but it doesn’t change that it’s the photo that’s the data leak, and not the tool. The tool just makes it vastly more accessible, and part of the task easier for curious human.
The tweet: tweet (Is the preview working for you? For me, it’s not).
The game is called geoguessing and those who do this regularly are crazy good at it, taking into account the kind of trees you see, where the sun and shadows are, even the color of the dirt and the pavement.
tragic that it doesnt include a gguf or safetensors file for easy access. ill load it up eventually. this would be very useful for invasive animal research
Add comment