Browsers have an absolute insane level of relatively unchecked permissions to do whatever they want on a client.

There's a lot of effort by browser developers to scope creep the browser into essentially being an OS-agnostic tech stack (one where, conveniently, code can be shipped across the network "as necessary", removing a lot of user agency for the software being ran); Chrome being the biggest driver of this, while Firefox has an extremely weak spine in trying to limit it.

It's fairly dire and I wouldn't be surprised if there's a lot more of these side channel attacks in a lot of web APIs.

Flash ended up getting blocked/banned by all browsers because it turned into a giant gaping security hole.

> By January 2021, all major browsers were blocking all Flash content unconditionally.

It looks like we-the-users need to be blocking any and every one of these parasites.

https://en.wikipedia.org/wiki/Adobe_Flash

I have a feeling they may have pushed for that more because it was controlled by a third party, and not the browser developers themselves.

Now that we have AI, can we go back to real apps and native tech stacks? And revert the browser to a text-display interface?

Unfortunately, real apps and native tech stacks can not only write data to your SSD, they can usually write data to the user directory however they want and they can read it as well!

Browsers are at least somewhat sandboxed

But you need relatively few native apps with those capabilities, in contrast to visiting many different websites for content consumption.

It very much applies to macOS, or do you know of a way to know what permissions a sideloaded macOS application will have before opening it that's accessible to regular users?

The very fact that you've qualified your question with "sideloaded" suggests that you are already aware that a non-sideloaded MacOS app is installed into a sandbox that is much more secure than anything available on a standard Linux desktop excepting possibly Qubes and Secureblue, and hardly anyone uses Qubes or Secureblue -- probably for very good reasons.

Yes, and I'm also aware that most macOS apps are still only available as a sideload, where sandboxing is optional and importantly not user visible before the app runs.

Why not sandbox the process then?

I don't know, maybe something about backwards compatibility, maybe nobody can agree on how to do it correctly. It hasn't happened for decades, so I'm not going to hold my breath.

Sure, can you propose an equally mature and battle-tested sandbox, ideally one with multiple implementations on a wide variety of OSes?

> can we go back to real apps and native tech stacks

Please God, no. If you're worried about the invasiveness of browser-based apps, native is out of the frying pan and into the fire

Except you’re not going to install native apps for the vast majority of things you use a browser for. You’re going to use the browser for content consumption and native apps for a few things that need system access.

Are you arguing that it's better to install a small handful of highly privileged applications as opposed to none?

Sure, once we have equivalent strong sandboxing for third-party apps I'm down to install random ones rather than visiting a website.

It's also the technology that will allow software to run without a continuous connection to the server. If you want to break out of a world where companies own your data it's the tech that is needed.

The uncomfortable part is that each step is usually justified by a real use case

My shortcut for launching "clean" Chromium session is `chromium --user-data-dir=$(mktemp -d)` -- each launch creates a new transient profile directory under /tmp, which is itself a RAM disk. Persistent settings are achieved by setting system-wide defaults in /etc/chromium, including using system-wide managed policy JSON.

Does this maintain your browser extensions (and their settings)?

Yes, extensions can be installed into the system-wide config via entries in the manged policy JSON. Settings configured in a specific browser session naturally won't persist, though, but defaults set in an initial preferences config will be present.

Is this surprising? Websites have long been silently writing to disk, for cache, cookies, and blobs. OPFS just provides a file-system-like API for ultimately the same functionality

Yes? From the paper:

"On Chrome and Safari, OPFS supports very large files, up to 60 % of disk space, which is more than sufficient to avoid the page cache on most typical systems, as even a small disk size of 64 GB would allow us to create a 38.4 GB OPFS file."

I am indeed surprised to learn that a random website can write a file that takes up 60% of my disk. Is this obviously a capability of Web browsers?

Not only that, but they don't even provide any visibility into what's being stored. Firefox developer tools doesn't even have OPFS browser functionality. IIRC I even saw some stuff about going out of the way to make it inaccessible by the user.

> Is this obviously a capability of Web browsers?

The main capability is RCE, but it seems that they need a way to store the payload.

There's a whole trend with websites not uploading anything to their servers due to privacy and whatnot, where do you suppose the data is being saved for repeat visits...

What your'e describing I would expect to be measured in kilobytes, not tens of gigabytes.

There is no reason for any person to think that a website needs to store data sized in the "full Ubuntu install" range to facilitate repeat visits.

You make a reasonable point, while kilobytes might be too little it probably shouldn't be 30gb. 5gb might be ok. In the settings it should be possible for the user to set their own limit. I am not familiar with browser storage but there is hopefully a mechanism to inform the user that their limit might not be enough.

> There is no reason for any person to think that a website needs to store data sized in the "full Ubuntu install" range to facilitate repeat visits.

Do you think people expect that for apps they've installed? Should those also be limited to a few MB?

I think that when I install an app, typically it will tell me up front how much disk space is required. For example, in the Mac App Store, the size of the app is at the top of the page.

> Should those also be limited to a few MB?

I also want to highlight that many/most websites that think of themselves as apps are at odds with their users in that perception. As an engineer, I know full well that e.g. the URL https://homedepot.com is powered by a sophisticated set of apps. But most users think it's a website.

This is important because people do & should have a different relationship with software they have chosen to install on their machines and websites. Yes, I know e.g. Figma.com does complex client-side stuff. Every Figma user would click a dialog to grant permission for it to do what it needs to do.

The problem is the current state of the art is that literally any website can spam up your disk and you don't even know. If I visit the website for a local radio station, or an e-commerce site, or university, or a site that will tell me what time it is now in a different time zone...I do not expect that it will download tons of data and store it on my disk in case I come back. That some engineers think that is reasonable is why the browser sandboxes need reinforcement.

> But most users think it's a website.

I'd be willing to bet most (younger?) users don't know what the difference between an app and a website is. Can't really blame them when the line between them has been blurring more and more over time. I think a growing number of users wouldn't even mention installing as a differentiator because you install PWAs (except on iOS).

Anyway, I think an important bit of information that was lost here is that browsers automatically purge data so that your disk doesn't fill up. If you're running low on space it will clean it up for you to make room.

> I do not expect that it will download tons of data and store it on my disk in case I come back. That some engineers think that is reasonable is why the browser sandboxes need reinforcement.

It's a reasonable thing to do for an app, so why not a website/PWA? Video games are a pretty good example where some stream assets while you are playing so that you don't have long install/update times. Getting in game faster is more important.

> most (younger?) users don't know what the difference between an app and a website is

Yeah, I doubt this because younger users overlap most heavily with "had to ask a parent for permission to install an app."

> It's a reasonable thing to do for an app, so why not a website/PWA?

Again, I chose to install an app and it has transparent install requirements. A website does not clear either of those bars.

Ten movies streaming across that, that Internet, and what happens to your own personal Internet? I just the other day got... an Internet [email] was sent by my staff at 10 o'clock in the morning on Friday. I got it yesterday [Tuesday]. Why? Because it got tangled up with all these things going on the Internet commercially. [...] They want to deliver vast amounts of information over the Internet. And again, the Internet is not something that you just dump something on. It's not a big truck. It's a series of tubes. And if you don't understand, those tubes can be filled and if they are filled, when you put your message in, it gets in line and it's going to be delayed by anyone that puts into that tube enormous amounts of material, enormous amounts of material.

That surprised me as well.

I thought the whole point of cookies, local storage, session storage, and indexed DB were to avoid what origin private file system is doing.

You mean I could have just saved stuff as a file this whole time instead of serializing it to a string? Why didn't we just do this from the start?

It's still sandboxed and deleted when the user clears private data for the website.

The main advantage it has over things like cookies, local storage, etc. is that it provides a byte-oriented, random access API and as a result, you can use third-party libraries like SQLite that expect a file API. Which is more important now that we have tools like Emscripten and WebAssembly that let you use existing C libraries on the web. At the same time it has security guarantees such that webpages cannot write arbitrary files that will be viewed and executed by the user.

Also, in theory you could use this side-channel attack on localStorage and sessionStorage. Its only requirement is that it needs an API that writes to disk where you can measure the latency of a synchronous call, since the fingerprinting is just measuring the interference pattern between disk accesses the attacking website does vs. disk accesses that other websites do.

Hostile LLMs? In my browser? At this time of the year?

> Wait, wait, wait: browsers allow websites to store junk on my drive?

Technically even a cookie is junk on your drive

> Without even asking whether the site can use local storage?

Would it be practical to ask permission for every site you visit? It would be better to periodically check the size of your home folder (where the browser profiles normally reside)

The funny part is that "put your browser profile on a ramdisk" used to sound like an obsessive performance tweak, and now it starts to look like a privacy mitigation

If you open an incognito window in chromium it is profile on ram

> Without even asking whether the site can use local storage?

Where did you see this in the article? I had some recollection that Firefox at least did require asking the user.

Firefox doesn't ask permission just to use localstorage, no modern browser does this. The closest thing you get is when a site wants to persist storage with "navigator.storage.persist()", which should prompt you for permission. But localstorage data usually persists anyway, and only gets deleted if the browser's storage is "under pressure", so I've never personally worked on a site or web app that had to use that API.

Btw. as per EU law (GDPR) website owners are required to aquire informed consent for any kind of client side storage if it contains information that is personal. And it has been ruled that any information that can be used to identify returning users is such.

People think the GDPR is just about cookies, but it is agnostic of the technology used.

Maximum fines: €20 million, or 4% of the company's total worldwide annual turnover of the preceding financial year — whichever is higher.

And informed consent means they need to know what data you collect/store for which purposes and there needs to be an equally easy to select No-Option.

This doesn't really address the issue here. The condition here is that a site might decide that it needs to store (say) a copy of the Red Hat server installation package on each user's local machine (20GB) to facilitate repeat visits.

The stored data is not related to the user at all. The problem is that the website gets to silently write 20GB to the user's disk.

You mean by default or it cannot be configured that way? I believe, I had Chrome configured to not allow storage by default, only for sites I added to an exclusion list. I cant remember now, but isnt there also an option to change the default on Firefox to deny or always ask for permission?

Just by default - I didn't know you could configure your browser to disallow storage by default.

I don't think LocalStorage allows you to store gigs of data though, and IIRC this method depended on the Origin-Private File System API.

And Web Developers want more and more OS features built into the browser. This is why I'm against it. Features are only ever abused.