Sorry I just read the GrapheneOS thread on the F-Droid signature pinning issue (the same issue I linked in the last paragraph of my first comment in fact), and I just wanted to add some comments. While I agree with most of the discussion there, the problem is that the alternatives are worse. Obtainium just pulls binaries directly from Github, where developer accounts have been compromised before. The Play Store has tons of malicious apps.

One of the main benefits of F-Droid is that they have standards. If you get an app from the default F-Droid repo, you can be reasonably certain that it is open-source and private. There are many apps like Bitwarden that couldn’t get included, and when you read the F-Droid Gitlab discussions on why, there are always good reasons. F-Droid will also warn you about telemetry and tracking, even if the app makes it into the default repo. These are things that Obtainium or the Play Store simply don’t provide.

The official GrapheneOS account wrote:

F-Droid automatically downloads and builds code, so despite their false marketing it does not protect users from the developers in any real way

Yes this does protect users. As I’ve mentioned before, it’s all too common for developers to sell their project to malicious third parties (often happens for browser extensions), or for developer accounts to be compromised (often happens for software packages, like NPM or PyPI). In these cases, the attackers will almost always change the pre-compiled binaries without updating the published source code. The only way to defend against this is via reproducible builds. F-Droid has been pushing for this, and the number of apps supporting reproducible builds has been growing year by year. Still, even without reproducible builds, I would rather trust F-Droid to protect their signing keys and accounts, rather than trust every app developer to do the same. After all, it only takes one compromised developer to compromise your phone.

Lastly, in the same comment by the GrapheneOS account, they said

If you continue to misrepresent, downplay and deny the many real issues about F-Droid, you’ll no longer be participating in our community.

This is very worrying to me, and makes me wary to participate in their community in the first place. As I just explained above I don’t agree with their logic, and now I see that this person is flaunting the fact that they can ban people for whatever they consider “misrepresentation”? I hope that the GrapheneOS community will recognize the dangers of centralizing all moderation power to somebody who seems so self-righteous.

Anyways I just wanted to share my thoughts on the thread, but thanks for the discussion as well, I bookmarked a lot of the links you shared and will be sharing them in the future!

secureblue includes modified images of CoreOS called securecore. While this doesn’t fix the issue you described, it is worth mentioning as a (technically) more secure option than both Debian and CoreOS.

Honestly I would not recommend securecore or secureblue for security. Small team, no track record, very little funding. I doubt their patches are audited by third-parties, and their userbase is probably so small that bugs are not found quickly. I’m sure you’ve already seen this PrivacyGuides thread on secureblue but the project is still very unstable. Their ideas may sound nice in theory, but patches can end up introduces more vulnerabilities than they fix. There are going to be breakages, changes in recommendations, bugs, regressions, and all of these impact security. I would not recommend it until their userbase is larger. You might ask how their userbase could ever get larger by my logic, which is why I’ll say that I’ll only recommend it for users who care about contributing and supporting the project, and improving the security of the future, even if it means sacrificing a bit of their own security at the present.

From my experience, having a large userbase and strong track record are the most reliable indicators for good security. You can always find articles criticizing old projects for security issues, but that’s simply because new projects aren’t under the same scrutiny (GrapheneOS is a rare exception). This is why I recommend Fedora Workstation/Silverblue over secureblue, Debian over CoreOS and securecore, and F-Droid over Accrescent. Though if you want to fight for a better future and test drive the hot new stuff, all the power to you.

I get what you’re saying, but at the same time if every developer released software as pre-compiled binaries on their website, installing stuff on Linux would become such a PITA. (This is different from how Windows works because apps for Windows are distributed using installers like xxx.msi, and Linux does not have a unified installation system across distros)

I’ve had the opposite experience, and started using Flatpaks after running into dependency conflicts once or twice when updating my system. Though I admit I’ve run into bugs with Flatpaks as well, just nothing as painful as a dependency conflict.

It’s impossible to know for sure whether you are tracked or not, but even the most basic fingerprinting mechanisms check browser version, and Reddit has advanced fingerprinting mechanisms to detect ban evasion. Couple that with the fact that 90% of my searches led me to Reddit, and it’s easy to conclude that Reddit correlated all my visits using my fingerprint, and thus has a history of all the things I have searched and been interested in for the past year, and sold that to Google. And Google has enough data on me from back when I used to use Google services, that they were probably able to link that activity to my real identity.

It sounds as though you were aware of this bug already. How did you find out? Did you notice it yourself or was there a notification somewhere?

You are right I should have linked directly to the workaround, sorry. Glad you got it sorted out though.

the point of tor is not to avoid fingerprinting, it’s to blend in

Fingerprinting and blending in are the same thing. You can’t blend in if you have a unique fingerprint. The Tor Project goes to great lengths to mitigate fingerprinting using their custom browser, it’s one of their main goals. It’s pointless to use Tor with a regular browser that doesn’t have those protections, because websites can just identify you by your fingerprint even when you are obfuscating your IP using Tor.

You are no more tracked by Reddit than you would be with up to date tor

Browser version is a major part of your fingerprint. It’s in your user agent, but that can be faked so there are additional mechanisms that check what javascript features your browser supports to get a more reliable read of your browser version. Use https://coveryourtracks.eff.org/ to learn more.

And fingerprinting is not a hack or exploit. It’s something that websites use for tracking, just like cookies. And I’m almost certain that Reddit fingerprints users to detect ban evasions.

Technically rollbacks are possible using regular packages, but in practice multiple packages will share dependencies and prevent you from downgrading just one of them. This is why it’s important that Flatpaks isolate dependencies between apps.

Are you saying that this bug would have been reported there? I don’t think I ever saw it, and I honestly doubt it was ever posted there. Unless you’re talking about the browser update announcements, but I would still need to check the Help > About page of my browser to notice that it didn’t match the latest version. As mentioned in my post, the Flatpak was updating like usual, the updates just weren’t affecting the browser.

Really, the main reason I made the post was to see if anybody else was affected, and see how other people avoided the bug. And aside from one other user, it really seems like nobody else was affected, which is surprising to me. The only reasons I can come up with are:

1. nobody installs Tor Browser using the Flatpak
2. everybody manually checks their browser versions
3. everybody installed or re-installed Tor Browser within the last year

Based on the comments I suspect #1 is the main cause. Which makes me lose trust in Flatpaks quite a bit. After all, if nobody is using them, then maintainers have less incentive to maintain them, and the worse they get.

Wow nice. Still not really friendly to beginners, since this is something they would have to dig into documentation to find, but it’s good to know

switch to the old.reddit.com site (onion version tends to work more often), and if that doesn’t work, switch Tor circuits (the option is under Tor Browser menu bar, I have it pinned to the top-bar for convenience)

Not to mention:

• better isolation between apps, no dependency conflicts
• ability to rollback to previous versions
• easily set environment variables and other launch options persistently
• transactional updates so if something weird happens during an update, the flatpak won’t be left in a corrupted state

This was an official Flatpak from Tor Browser, so there’s no reason why it should be less reliable than the packages from distribution maintainers. Not to mention for atomic distros, flatpaks are the official way to install software.

You can check the Tor Project blog to figure out the latest release, and go to your Tor Browser’s menu > Help > About Tor Browser to see if it matches. It should be version 14.0.7. If it is not, the fix is detailed in the Github issue I linked in the post

This seems like something that Flatpak should be able to handle though. Afaik Mullvad Browser never had this issue. Flatpaks also have numerous advantages, like automatically handling desktop shortcuts.

I hope so, Flatpaks are becoming the default way of installing packages, especially with the rise of atomic distros.

Done, reposted to linux@lemmy.ml and privacy@lemmy.dbzer0.com. Though maybe linux@lemmy.ml was unnecessary because this post is already on the lemmy.ml instance…

First off, props on the detailed and informative post. I’ve never seen a post so packed with links and citations. I’d just like to share some of my own experience:

In regards to Debian vs atomic distros. First off, most recommendations for Debian are recommending it for use on the server. I definitely agree that on the desktop, you are better off with a more up-to-date distro, especially for browser patches. But for the server, after having used both Debian and Fedora CoreOS (an atomic distro for servers) for over a year each, I trust Debian more in terms of security and stability. For example, last summer when there was a major OpenSSH vulnerability, Debian had already patched it, because the security researchers had notified the Debian maintainers prior to the announcement. CoreOS on the other hand, took multiple weeks to release the fix. I also ran into some coredumps on Fedora CoreOS. It was only once or twice, but I never experienced the same on Debian. The main reason why I trust Debian is simply because it’s an industry standard. Billions if not trillions of dollars are on the line if Debian is compromised. CoreOS and atomic distros are just not popular enough to receive nearly as much attention. There’s safety in numbers. That’s why for the server, I’d recommend Debian, while for the desktop, Ubuntu or Fedora are better choices. Though if you really want security on the server, I would recommend Proxmox, which uses a similar security model as Qubes. Note that Proxmox is based on Debian.

As for the topic of F-Droid, you brought up the PrivSec article on F-droid security issues. This article is a few years old and is always brought up in criticisms against F-Droid. My main problem with it is that it downplays the importance of open source. One thing not mentioned in the article is that ideally, you shouldn’t even need to trust the developer. That’s one of the benefits of open source. Those familiar with the world of browser extensions are also all too familiar with how often the developer sells the project to a malicious party, who can then backdoor the published extension without updating the source code. Now, open source is only secure if it’s audited, something you mentioned in your post, but in my experience just the fact that it can be audited is good enough to scare away bad actors. Afaik F-Droid has had zero malware. Despite being a small store, that’s still extremely impressive, and speaks for itself. There is still the danger that F-Droid itself is compromised, but that can be solved with reproducible builds, which is something the Play Store can’t offer due to Play App Signing, while F-Droid is pushing for it.

Though that is just in theory. I should mention that there was a pretty worrying issue found in F-Droid reproducible builds recently. I still trust the security of F-Droid more than the Play Store though.