I am a long-time NoScript extension (https://noscript.net/) user. For those who don’t know this automatically blocks any javascript and let you accept them (temporarily or permanently) based on the scripts’ origin domain.
NoScript as some quality-of-life option like ‘accepting script from current page’s domain by default’ so only 3rd parties would be blocked (usefull in mobile where it is tedious to go to the menu).
When I saw LibreJS (https://www.gnu.org/software/librejs/) I though that would be a better version of NoScript but it is quiet different in usage and cares about license and not open-source code (maybe it can’t).
Am I the only one who thought about checking for open-source JS scripts filtering (at least by default)? This would require reproducibility of ‘compilation’/packaging. I think with lock files (npm, yarn, etc) this could be doable and we could have some automatic checks for code.
Maybe the trust system for who checks could be a problem. I wanted to discuss this matter for a while.
Thanks for your answer.
First I don’t even grasp what a “service owner” is.
Second, for JS front-end openness there are already a bunch of app (web, android) that are open-source and secured. Everything has dependencies nowadays, this doesn’t prevent good security. Think all the python app and their dependencies, rust, android… even c\c++ packages are built with dependencies and security updates are necessary (bash had security issues).
I think with JS scripts it’s actually even easier to have good security because the app is ran in our web browser so the only possible attacker is the website we are visiting itself. If they are malicious then the close-sourced JS script is even worse. Unless you count 3rd party scripts embedded that bad dev uses in their website without even thinking about trusting them. That is also awful in both open or close source environment.
So even having imperfect security (which happens regardless to openness), who is the attacker here? I would rather run js script on my end if the code can be checked.
The people who build & run the software & servers that serve the website, who amongst other things have an interest in keeping the service available, secure, performant, etc.
Particularly with laws like GDPR, these services owners are motivated to be as secure as practically possible otherwise they could receive a bankrupting fine should they end up leaking someone’s data. You’ll never be able to convince anyone to lower the security of their threat model for that reason alone, before anything else.
The code published and the code running on a server cannot be treated as equivalent for several reasons, but here’s two big ones:
Firstly, there’s the similar issue as with compiled binaries in other languages: it’s tough (or impossible) to verify that the code published is the same code that’s running. Secondly the bundled and minified versions of websites are rarely published anyway, at most you get the constituent code and a dependency list for something completely open source. This is the bit I referred to before as trying to untoast bread, the browser gets a bundle that can’t practically be reversed back into that list of parts and dependencies in a general purpose way. You’d need the whole picture to be able to do any kind of filtering here.
The only possible attacker is not the website itself (though it’s a lot more limited if the site implements CSP & SRI, as mentioned in my other comment). XSS is a whole category of attacks which leverage an otherwise trusted site to do something malicious, this is one of the main reasons you would run something like noscript.
There have also been several instances in recent years of people contributing to (or outright taking over) previously trusted open source projects and sneaking in something malicious. This then gets executed and/or bundled during development in anything that uses it and updates to the compromised version before people find the vulnerability.
Finally there are network level attacks which thankfully are a lot less common these days due to HTTPS adoption (and to be a broken record, CSP & SRI), but if you happen to use public WiFi, there’s a whole heap of ways a malicious actor can mess with what your browser ultimately loads.
OK I got it, you are completely out of the loop here.
You do not grasp the idea of NoScript and other JS filtering extension. This is not about server code, your all arguments is baseless here.
By the way JS refered to Javascript and not NodeJS.
Anyway I got you whole company/business talk about “keeping the service available, secure, performant” and “GDPR […] bankrupting fine”… yeah lemmy.world.
I’m a full-stack software developer working in the financial sector, their statement is factual.
Companies will never want to take on liability that has the potential to bankrupt them. It is in their best interest to not reveal the version of libraries they are using as some versions may have publicly known vulnerabilities, and it would make it incredibly easy for attackers to build an exploit chain if they knew the exact versions being used.
Securing client code is just as important as securing server code, as you don’t want to expose your users to potential XSS attacks that could affect the way the page gets displayed, or worse, leak their credentials to a third party. If this happened in the EU or some parts of Canada, and it’s been found that the company reduced their threat model “for the sake of openness”, they would likely be fined into bankruptcy or forced to leave the market.
Unfortunately, this is one of those cases where your interests and ethics will never be aligned with those of service owners as they are held to a certain standard by privacy laws and other regulations.
No need to get aggravated, I completely grasp it, you’ve possibly misunderstood or not entirely read my comment if that’s your takeaway.
I’m not talking about server code specifically, I’m going through the stages between the source code repo(s) and what your browser ends up receiving when you request a site.
NodeJS is relevant here because it’s what runs nearly all major JS bundlers (webpack, vite, etc), which are what produces the code that ultimately runs in the browser for most websites you use. Essentially in a mathematical sense, the full set of dependencies for that process are a part of the input to the function that outputs the JS bundle(s).
I’m not really sure what you mean with that last part, really, anyone hosting something on the internet has to care about that stuff, not just businesses. GDPR can target individuals just as easily as for-profit companies, it’s about the safety of the data, not who has it—I’m assuming you would not want to go personally bankrupt due to a deliberate neglect of security? Similarly, if you have a website that doesn’t hit the performance NFRs that search engines set, no one will ever find it in search results because it’ll be down on page 100. You will not be visiting websites which don’t care about this stuff.
Either way, all of that is wider reasoning for the main point which we’re getting away from a bit, so I’ll try to summarise as best I can:
Basically unless you intend your idea to only work on entirely open source websites (which comprise a tiny percentage of the web), you’re going to have to contend with these JS bundles, which as I’ve gone into, is basically an insurmountable task due to not having the complete set of inputs.
If you do only intend it to work with those completely open source websites, then crack on, I guess. There’s still what looks to me like a crazy amount of things to figure out in order to create a filter that won’t be able to work with nearly all web traffic, but if that’s still worth it to you, then don’t let me convince you otherwise.
Edit: typo