In this case, I agree that it’s a low priority patch. Here’s what you must do as an attacker. Decide for yourself whether it sounds practical for general deployment.
Requirements: Fill OPFS storage with an arbitrarily large amount of data which at least exceeds RAM, but may require up to 60% of SSD, then lock up a thread with random reads while a worker thread hosts a model that you feed any detected latency clusters.
Even if users don’t notice their fans maxing / battery burning / memory+storage disappearing and kill the tab themselves, this definitely will be the first tab offloaded by most browsers and OSes shortly after it is sent to the background.
That means you have a brief window where you might get the chance to guess which sites a user is visiting. Your guess is likely far less than 89% accurate (PoCs illustrate in optimal conditions where models are often deliberately overfit to specific machine(s) and locale) outside a hyper-targeted attack, you will be lucky for coin toss levels of certainty for any guess.
Is this an attractive attack vector?
The actual paper (PDF) this is based on gives much better information than the article. From that we get some really key information:
To allow FROST to measure SSD contention, the victim must perform activities that result in storage accesses to the same disk as the file used for contention measurement
This can’t ready your SSD. It can only listen in on the conversation between your CPU and SSD when something else reads it or writes to it. The whole FROST approach has a number of clever tricks to generate reads from open applications though. Further, it requires the attacker’s code to be running in an active browser session.
Also, If you have two SSDs, and your browser is on one, this FROST approach can’t see anything written to or read from on the other SSD.
Lastly, there’s a mention in the paper about hardware based SSD encryption being vulnerable, there’s no mention of Software Whole Disk Encryption. Given how the researchers are using the SSD timing exploit, I would guess that a software (not hardware) whole disk encryption might be immune to this attack because the patterns of timings would be different with encrypted data being written to the SSD (instead of the data being encrypted by the SSD when written.
The paper also mentions that it also takes downloading a 1GB OFPS file and JS in use.
This isn’t so much “researchers can track you” so much as “it’s theocratically possible with stock laptops and browsers, within limitations.”
Sounds like a bunch of timing attacks could be rendered useless if access to an accurate timer required special permission. And without the permission, it either limited the resolution or added random jitter to any timer APIs.
The attack creates a large OPFS file on the victim’s SSD, with both Chrome and Safari allowing a website to claim up to 60% of total disk space through OPFS, which on a 256GB drive is over 150GB.
Am I reading this right? 60% of all your disk space can be confiscated by some random web site? I gotta figure out how to get my browser cache onto some tiny partition.
I’ve done it with some apps/games by placing the folder in question on a separate drive/partition and using junction points (I use Junction Link Magic, but you can do it manually from command prompt) to basically create a ghost of the folder in the original location that routes everything to the new location.
You could create a small hidden partition just for the browser cache folder to reside on using this method.
Replicating this on Linux would be as simple as ln -s to make a symbolic link
Symbolic links for the win!
I wonder if, at any point, anyone stopped to ask themselves, “did I really go to school just so I can ply my knowledge and expertise to find even more ways to fucking track people who expressly don’t want to be tracked so we can use the data for ad revenue (if not for other, even worse things)”?
software dev here.
I worked with a guy who was implementing application monitoring for clientside applications. think of it like google analytics for single page apps. he proposed we could require users install a browser plugin to make it easier to track and monitor the users on our app with the added benefit we could track them on other websites like our competition.
one person in a room of about 11 people spoke up about the implications of privacy and the backlash we might have from our user base when they find out that we basically just installed a keylogger in their browser.
the only thing that stopped this plan from going forward was the risk of losing users and potential revenue loss.
my point in all this is to answer your question. no, most people have stopped thinking about their actions and are just creating “solutions” to problems that don’t exist.
Hey I’ve been in that room! I don’t get it, I can’t live with that for of thing. And this is why I only have like 2 or 3 extensions (all ad blockers).
Execs love this shit. I only had one exec who pushed not to do that or open Pandora’s box.
He made a ton of cash, cashed out, and retired at 30 something. Awesome dude, I miss working under him.
Out of curiosity, why 2-3 different ad blockers?
I have similar.
- privacy badger
- ublock
- adblock plus
and have pihole on my network.
They often do different things.
Privacy badger, ublock, https everywhere are just required tools.
It’s wild how quickly morality falls to the wayside (and is subsequently paved over). Especially crazy to abandon one’s moral standing early on the path of solving problems that don’t exist to appease people who don’t care for a chance at the advancement of a career that you can’t take with you in a field that could be wiped out by a solar flare, all to end up making the world a worse place for subsequent generations (I’m not a bleeding-heart idealist, lol).
I often think about a few people I know who have psych degrees. All were told, in different years, that if they wanted to make money as a psychologist, they needed to get in with tech companies. Some even got job offers.
I studied data mining (now machine learning) and statistics.
I’ve spent my career explicitly NOT plying my knowledge this way. I don’t know how people do it.
I’d say my deep knowledge on how to track people has made me pretty averse to a lot of online things.
You know you can build marketing attribution systems and advertising metrics without violating user privacy.
But advertisers really like the idea of invading privacy and they pay out the nose for it.
Good on you. Few are willing to take the overgrown path. And, funny how people who work with the subject matter often avoid it- the cybersecurity guy who doesn’t own a computer, the guy who services food processing equipment who refuses to buy premade food, the guy who works/ed for the DoD who doesn’t own a phone, etc.
Would you mind sharing some of the online things you’re averse to, besides all that is implied by being on the Fediverse? I’m still new to this stuff.
I work IT adjacent in physical security systems (cameras, access control, intrusion systems etc.). Everyone looks confused when they ask what I have at home or what they should install and I tell them fuck all of this surveillance state/must know every time someone thinks about my house bullshit. I push back on a lot of corporate garbage as well and I’m lucky enough to work off a company that listens and balances security with privacy when I steer us that way.
I think this is pretty common in tech fields.
I’ve got the same thing. I had someone ask me what I do for backups and they thought I was joking when I told them I have a good printer. They couldn’t get their head around the idea that I don’t even have a home network to attach a NAS to, and thought I was just being condescending. I had a similar conversation when asked how to secure an Alexa.
Just things that can be correlated. Time, device, network, accounts, and apps all correlate. Precise location, device sensors, etc also correlate.
You have to decide what you want security or privacy against, then you have to be mindful always.
Every internet connection is a fingerprint.
E.g. The second you use that device on an VPN all your apps phoning home, checking notifications, logging events, etc. collapse your profile and deanonymize your anonymous activity.
So I actually use a dedicated device for anything I want a VPN on.
Opsec almost requires that you need a public device for your regular use, and a secondary device with limited scope, third party OS for higher privacy for anything you actually don’t want to share.
It’s safer to tunnel specific whitelisted connections through a VPN than whole device VPN for that reason (the less traffic goes to VPN the better). iOS VPN doesn’t work for that reason.
If you want VPN security, the best way is to run a container with only VPN networking, then a second container with the service you want protected and route all networking through the VPN container.
Also, say no to Chrome based apps looking for devices on your network. That uniquely fingerprints you across tons of surfaces.
They say it’s for chrome cast or something but it’s too much info to share.
That’s really great, thank you. I’ve got a working knowledge of applying opsec and related principles, but my understanding quickly drops off when we get into the why. That’s super helpful.
It is really hard to sort through job listings using ethics as your criteria.
Bet
“Google says fingerprinting is not a security vulnerability”. That is a very google thing to say.
They really gave internet browsers too much access. Why the fuck does my browser need this level of clearance
And this is why you shouldn’t use any browser based on Chromium, because where Google goes (implementing ever more invasive APIs) they all follow.
They better hope there isn’t another pandemic. Cuz I am officially labeling that quantity of free time as “1 Degoogle”. and if I ever get my hands on another fuckin degoogle I am going to degoogle all over my fucking house
Holy shit, what do you mean all over your house? How did you ever let it get that bad?
You can Degoogle one step at a time. Make it a weekly habit to spend 10 minutes Degoogling every week and you won’t need to wait for another worldwide shutdown
True. And I did hear if you degoogle at least 20 times a month it reduces your risk of prostate cancer.
deleted by creator
Becau of the push for web apps to get around platform (and platform store) limitations.
e.g. Apple banned apps for vapes (not just talking about nic vapes but e.g. there’s a number of cannabis flower vapourisers that use Bluetooth for fine tuned settings, those were forced to move over to web apps as the native apps simply got pulled), but also software like ESPHome is completely web based and needs access to raw USB devices to write the new firmware onto them, the list goes on.
Main issue seems to be that a lot of these APIs don’t require explicit user approval. USB, Bluetooth does, but apparently accessing detailed system statistics doesn’t? Make that make sense…
Well it’s all potential advertisement revenue
I’m not even sure I’d allow them to know my resolution…
Modern browsers are basically an OS of their own at this point. And, in fairness, for many users their actual OS is effectively just a means by which they access a browser. This is not even mentioning something like Chrome OS.
I’m not saying it’s a good idea or that it should be this way, but it is where we are.
If you don’t allow ReCaptcha access to your address book, the website will fail to load.
I looked it up - Firefox does not allow OFPS storage in private mode since November 2022 , so that is an option at least.
I wonder if librewolf has this option on by default
I don’t think so - from how i understand it, everything where you upload a file to a server to edit it there wouldn’t work at all.
The flow looks like this:
You select a file for upload - the browser creates a file in OPFS storage representing the original file - any change you make serverside are replicated to the copy in OPFS storage - when you save the file, you don’t actually have to download it, but the file gets moved from OPFS to wherever you save the file. This prevents long downloads and a lot of warnings and the file in OPFS storage is encrypted because of HTTPS.it’s explained in detail here: https://web.dev/articles/origin-private-file-system
But only in private mode?
I remember when browsers just showed text.
We should just throw away the web and do something new. Maybe Fidonet over Reticulum so we can use radio.
So that’s not what the paper says
So the file has to exceed available RAM to benchmark the SSD performance? How viable is that at all? You’d be downloading gigabytes.
You don’t download the file. The JavaScript generates the file right on disk.
Ah that makes more sense. Seems like something easy to detect at least.
It’s been a while but doesn’t Windows let you know when you exceed RAM usage and hit paging file?
You didn’t hit the page file. This is OPFS, an in-browser filesystem that is sandboxed to each origin (essentially to each website), not directly accessible by the user, and exempt from the security checks that would guard access to the regular filesystem.
Yeah, that sounds to me like it needs a major revision.
but in order for the file to use all available RAM, other processes that still need memory will eventually trigger the out of memory warning… no?
unless I’m completely misunderstanding and OPFS has a set limit of RAM usage before it automatically starts writing to drives.
You also have to provide access to your computer so the attacker can produce labeled training data for the neural network that performs the pattern matching for the actual fingerprinting.
Because that’s what they did in the paper: they got the data and performed the attack on the same machine. There’s no evidence presented in the paper that this identification could be generalised to arbitrary machines and configurations without prior access.
So yes, this is a complete nothingburger.
Good i dont allow every Javascript on websites (usually)
Fun fact: running inside a flatpak or snap in no way mitigates this 😪
deleted by creator
