If I can miss subtle and not-so-subtle bugs in my own team's code that I'm working on every single day, what are the chances I'll get it right on somebody else's project?
Being able to MITM any HTTPS request on my system is a privilege I'd not grant lightly. It's up there with browsers, browser extensions (with "all content on all sites" access), and password managers in terms of blast radius in case anything goes wrong.
No one will do this, and those that read source code during installation do not review it for every upgrade. It's one of those 'just do this!' arguments that has little to no basis in reality. There's more of them replying to the parent comment: "Just do this! Just compile a thing! Just verify signatures for every update!". Come on... Meanwhile the negatives immediately implicate anyone with access to the executable.
You don't know this person, and I see no personally identifiable information to make me trust them. They could literally be a state actor right now! We've also seen so many large supply-chain attacks over the last decade which could easily target a tiny project like this.
But doesn’t that apply to chromium / firefox as well (or any other big application). Web browsers are insanely huge, nobody is reading the entire code. What makes this different?
To take a different argument, it’s another party that’s less well known that you have to trust. The more parties you trust, the less secure things become. Whether that additional risk is something you want to take on is a personal decision. MITM SSL traffic would make me uncomfortable.
If you really wanted/NEEDED this, you could definitely go through all the code. It would take a bit, but it's doable with determination (lol). Also, you don't have to necessarily review all of the code every update. All you have to do is view the changes/new commits every time you want to update.
The hardest part is determining that you want to go through all of this hassle to replicate something browser extensions already do (for the most part).
If you believe you can find even just all unintentional bugs, let alone deliberate security vulnerabilities, you've never looked at the underhanded C contest [1].
> All you have to do is view the changes/new commits every time you want to update.
These can be thousands of lines of code per day in busy projects.
I'm as paranoid about this as you, but this type of verification seems easier today with AI tools. I'm not aware of any that do this, but if LLMs can give insight about what a piece of code is doing, they can surely be trained to detect possible suspicious behavior. Perhaps even by inspecting a binary, but certainly by processing code.
Maybe for well-intended code (and even there I have my doubts – the halting problem says hi!), but most definitely not for malicious backdoors at this point.
The build steps are provided as a GitHub action in the repo. You can audit the build pretty easily, or if you're super paranoid you can build it yourself by following the build steps.
Even if the app is trustworthy, it still adds an attack vector to your system. It could have a bug, or the certificate could be exploited by another program.
How do you know that the downloaded binaries are built by the GitHub action? If I must audit the code and build it myself every release, then how is this a usable product?
Not a product. It's literally free, free as in free speech, and free as in you're free not to use it.
Building the code yourself for every update is also a solved problem on every system with a feature complete package manager, including Windows. Trust is not so easily solvable, but if you trust nobody, you can choose to look at ads.
In fact, that's what I'm trying to say. The line we draw at "hey we can at least audit open source" is a fully imaginary one. It's a false comfort we create. It's the Kool-Aid we drink.
I don't have any trust in any of those components you mentioned, but I came to terms with the risks associated with using them as part of my threat model. However, I find the notion that open source is somewhat safer because "we can audit it" exaggerating if not misleading. It's not a valid argument, and it should never be used because there's no way to do it in an either practical or consistent way for the users of the said product.
That's true, but if it is trying to exfiltrate data you will probably see more packets coming out than going in. If the concern is that the SSL cert they add allows them to decrypt your data, and you see it sending your requests anywhere it shouldn't, you'll get a window into what they are trying to do. It's not perfect, but it is possible to gain information in such a way.
Unfortunately many people do like that, but I hate it and I don't do stuff like that (and it seems many other people hate it too; fortunately you can just download the file without immediately executing it, if you want to view (and possibly modify) it).