Not discrediting Open Source Software, but nothing is 100% safe.
Luckily there are people who do know, and we verify things for our own security and for the community as part of keeping Open Source projects healthy.
Open source software is safe because somebody knows how to audit it.
And to a large extent, there is automatic software that can audit things like dependencies. This software is also largely open source because hey, nobody’s perfect. But this only works when your source is available.
Except when people pull off shit like Heartbleed.
See my comment below for more of my thoughts on why I think heartbleed was an overwhelming success.
And you help make my point because openssl is a dependency which is easily discovered by software like dependabot and renovate. So when the next heartbleed happens, we can spread the fixes even more quickly.
Enterprise software inventory can unfortunately be quite chaotic, and understanding the exposure to this kind of vulnerability can take weeks if not longer.
Also because those people who can audit it don’t have a financial incentive to hide any flaws they find
It’s safe because there’s always a loud nerd who will make sure everyone knows if it sucks. They will make it their life mission
Will that nerd be heard or be buried under the scrutiny?
I’ll listen to them because I love OSS drama. But you’re right that they may just get passed over at large
My very obvious rebuttal: Shellshock was introduced into bash in 1989, and found in 2014. It was incredibly trivial to exploit and if you had shell, you had root perms, which is insane.
env x=‘() { :;}; echo vulnerable’ bash -c “echo this is a test”
Though one of the major issues is that people get comfortable with that idea and assume for every open source project there is some other good Samaritan auditing it
I would argue that even in that scenario it’s still better to have the source available than have it closed.
If nobody has bothered to audit it then the number of people affected by any flaws will likely be minimal anyway. And you can be proactive and audit it yourself or hire someone to before using it in anything critical.
If nobody can audit it that’s a whole different situation though. You pretty much have to assume it is compromised in that case because you have no way of knowing.
Oh definitely, I fully agree. It’s just a lot of people need to stop approaching open source with an immediate inherent level of trust that they wouldn’t normally give to closed source. It’s only really safer once you know it’s been audited.
Have you seen the dependency trees of projects in npm? I really doubt most packages are audited on a regular basis.
But eventually somebody will look and if they find something, they can just fork the code and remove anything malicious. Anyways, open source to me is not about security, but about the public “owning” the code. If code is public all can benefit from it and we don’t have to redo every single crappy little program until the end of time but can instead just use what is out there.
Especially if we are talking about software payed for by taxes. That stuff has to be out in the open (with exception for some high security stuff - I don’t expect them to open source the software used in a damn tank, a rocket or a fighter jet)Fun fact*: the software in the most advanced dildos come from old missile guidance systems the government isn’t using anymore.
*not a fact, but hopefully fun.
Maybe not a fact but I will still accept it as canon
No, missle.
Agreed.
this was indeed fun
Thanks 😁
Closed-source software is inherently predatory.
It doesn’t matter if you can read the code or not, the only options that respect your freedom are open source.
Do you know how to audit the code?
Yes?
I don’t. But I trust you
Just learn the basics and you don’t need to trust. Like… everything science.
Now audit the linux kernel
No.
Fair enough
LGTM
Let’s go to Mars?
Completely missing the point. Collective action is what makes open source software accessible to everybody.
You dont NEED to be able to audit yourself. Still safer than proprietary software every way you look at it.
While I generally agree, the project needs to be big enough that somebody looks through the code. I would argue Microsoft word is safer than some l small abandoned open source software from some Russian developer
no, proprietary software its always possible malware and you have no weapon against it. being able to audit is always better.
That’s true, but I’m not a programmer and on a GitHub project with 3 stars I can’t count on someone else doing it. (Of course this argument doesnt apply to big projects like libre office) With Microsoft I can at least trust that they will be in trouble or at least get bad press when doing something malicious.
I mean if a github project has only 3 stars, it means no one is using it. Why does safety matter here? Early adopting anything has risks.
This is kind of a false comparison. If it has 3 stars then it doesn’t even qualify for this conversation as literally no one is using it.
undefined> With Microsoft I can at least trust that they will be in trouble
lol yeah if anybody finds out… something something NSA
Ehmm. if nobody uses it, it kinda doen’t matter if it’s safe. And for this example: I bet more people had a look at the code of LibreOffice than MS Office. And i dont think it sends telemetry home in default settings.
I think they’re talking about onlyoffice.
no , but I know a bunch of passionate geek are doing it.
- Yes, I do it occasionally
- You don’t need to. If it’s open source, it’s open to billions of people. It only takes one finding a problem and reporting it to the world
- There are many more benefits to open source: a. It future proofs the program (many old software can’t run on current setups without modifications). Open source makes sure you can compile a program with more recent tooling and dependencies rather than rely on existing binaries with ancient tooling or dependencies b. Remove reliance on developer for packaging. This means a developer may only produce binaries for Linux, but I can take it and compile it for MacOS or Windows or a completely different architecture like ARM c. It means I can contribute features to the program if it wasn’t the developer’s priority. I can even fork it if the developer didn’t want to merge it into their branch.
Regarding point 2. I get what you’re saying but I instantly thought of Heartbleed. Arguably one of the most used examples of open source in the world, but primarily maintained by one single guy and it took 2 years for someone to notice the flaw.
Uhh… so? The NSA was sitting on the vulnerability for EternalBlue in Windows for over 5 years.
Dont understand what that has to do with the discussion so far. How is this relevant here?
No more or less relevant than heartbleed. Yes vulns exist in open source software, sometimes for a while. Being open source can lead to those vulns getting discovered and fixed quicker than with closed source.
And how does this negate my initial point that you shouldn’t trust in the security of something just because it is open source? I think you misunderstood what I was saying.
Nobody does that.
Alright then, have a nice day!
You shouldn’t automatically trust open source code just because its open source. There have been cases where something on github contains actual malicious code, but those are typically not very well known or don’t have very many eyes on it. But in general open source code has the potential to be more trustworthy especially if its very popular and has a lot of eyes on it.
It’s one reason I haven’t rushed to try out every lemmy app that has come out yet.
A lot of bad takes in here.
Here are a few things that apparently need to be stated:
- Any code that is distributed can be audited, closed or open source.
- It is easier to audit open source code because, well, you have the source code.
- Closed source software can still be audited using reverse engineering techniques such as static analysis (reading the disassembly) or dynamic analysis (using a debugger to walk through the assembly at runtime) or both.
- Examples of vulnerabilities published by independent researchers demonstrates 2 things: people are auditing open source software for security issues and people are in fact auditing closed source software for security issues
- Vulnerabilities published by independent researchers doesn’t demonstrate any of the wild claims many of you think they do.
- No software of a reasonable size is 100% secure. Closed or open doesn’t matter.
Very good points here, especially your last point
As you increase the complexity of a system, it makes sense that your chance of vulnerability increases. End of the day, open source or not, you will never beat basic algorithm principals and good coding practice.
I would however argue that just because closed source code is possibly reversed doesn’t mean it’s easier or as reliable as having the source code. As long as corporations have an interest in possession there will always be someone striving and spending ungodly amounts of money to keep their castle grounds gated heavily which makes securing them en mass much harder and slower
I agree, it takes longer to audit closed source software. Just wanted to point out it’s not impossible, as long as you have a binary.
Closed source software can still be audited using reverse engineering techniques such as static analysis (reading the disassembly) or dynamic analysis (using a debugger to walk through the assembly at runtime) or both.
How are you going to do that if it’s software-as-a-service?
See the first bullet point. I was referring to any code that is distributed.
Yeah, there’s no way to really audit code running on a remote server with the exception of fuzzing. Hell, even FOSS can’t be properly audited on a remote server because you kind of have to trust that they’re running the version of the source code they say they are.
Ohhh, code that is distributed. The implication of that word flew over my head lmao, thanks for the clarification.
You can always brute force the SSH login and take a look around yourself. If you leave an apology.txt file in /home, I’m sure the admin won’t mind.
Lol, unlikely SSH is exposed to the net. You’ll probably need an RCE in the service to pop a shell.
That’s not universally true, at least if you’re not on the same LAN. For example, most small-scale apps hosted on VPSs are typically configured with a public-facing SSH login.
Ah yes let’s audit through reverse engineering. The same thing huh? xD
Second bullet point, it’s much easier to audit when you have the source code. Just wanted to point out it’s not important to audit closed source software. It’s just more time consuming and fewer people have the skills to do so.
Also, just because you can see the source code does not mean it has been audited, and just because you cannot see the source code does not mean it has not been audited. A company has a lot more money to spend on hiring people and external teams to audit their code (without needing to reverse engineer it). More so than some single developer does for their OSS project, even if most of the internet relies on it (see openssl).
And just because a company has the money to spend on audits doesn’t mean they did, and even when they did, doesn’t mean they acted on the results. Moreover, just because code was audited doesn’t mean all of the security issues were identified.
Yup, all reasons why it does not matter if the software is open or closed as to how secure it might be. Both open and closed source code can be developed in a more or less secure fashion. Just because something could be done does not mean it has been done.
Nah I wouldn’t say that. Especially if you consider privacy a component to security. The fact that a piece of software can more easily be independently reviewed, either by you or the open source community at large, is something I value.
Good security is a component to privacy. But you can have good security with no privacy - that is the whole idea of a surveillance state (which IMO is a horrifying concept). Both are worth having, but my previous responses were only about the security aspect of OSS. There are many other good arguments to have about the benefits of OSS, but increased security is not a valid one.
The point is not that you can audit it yourself, it’s that SOMEBODY can audit it and then tell everybody about it. Only a single person needs to find an exploit and tell the community about it for that exploit to get closed.
Exactly! I wait on someone who isn’t an idiot like me to say, “ok, so here’s what’s up guys.”
“Transparent and accountable government is a waste of time because I personally don’t have the time to audit every last descision.”
OP, you are paranoid beyond belief.
deleted by creator
I really like the idea of open source software and use it as much as possible.
But another “problem” is that you don’t know if the compiled program you use is actually based on the open source code or if the developer merged it with some shady code no one knows about. Sure, you can compile by yourself. But who does that 😉?
But another “problem” is that you don’t know if the compiled program you use is actually based on the open source code or if the developer merged it with some shady code no one knows about.
Actually, there is a Debian project working on exactly that problem, called reproducible builds
yes and others are working on it, also! i believe some android folks are (f-droid iirc), and i’ve heard about it elsewhere. this stuff is super nerdy (so therefore cool to nerds such as myself). before the internet existed it would be so hard to even imagine the need for this sort of thing!
You can check it using the checksum. But who does that?
In all seriousness I am running NixOS right now using flakes. The package manager compiles everything unless a trusted source already has it compiled, in which case the package manager checks the checksum to ensure you still get the same result and downloads that instead. It also aims to be fully reproducible and with flakes it automatically pins all dependency versions so next time you build your configurations, you get the same result. It is all really cool, but I still don’t understand everything and I’m still learning it.
Based NixOS user
I love NixOS but I really wish it had some form of containerization by default for all packages like flatpak and I didn’t have to monkey with the config to install a package/change a setting. Other than that it is literally the perfect distro, every bit of my os config can be duplicated from a single git repo.
Great points. I kinda feel the same with containerization. I have been wanting change my OS on my home server and while NixOS is great for that, I have decided to do things differently and use OpenSUSE Micro OS. My plan was actually Fedora Core OS, but after that Red Hat drama I decided to run with SUSE instead. It is an immutable distro with atomic upgrades that is designed for being a container host. It uses Ignition as the configuration for setting up things like users, services, networking, etc. My plan is then to just use containers like I was doing before on Fedora Server and for the other things to use Nix to build container images. Instead of using DockerFile, you’d use Nix Flakes to create really minimal images. Instead of starting with a full distro like Alpine, Nix starts from scratch and copies all dependencies over as specified in your flake. So the image only contains the absolute minimum to run. I think I’d be a fun side project while learning more about Ignition, Linux containers and Nix Flakes.
As for your point on config, I think it’s just part of the trade offs of NixOS. You either have a system that can be modified easily at anytime through the shell or you have a system that you modify centrally and is fully reproducible. You can already install packages with nix-env in the command line without changing your config, but that also won’t be reproducible. Maybe a GUI app for managing your config and packages could be helpful, although I’m pretty sure that’s low priority for NixOS right now.
Many people do that…
We trust open source apps because nobody would add malicious codes in his app and then release the source code to public. It doesn’t matter if someone actually looks into it or not, but having the guts to publish the source codes alone brings a lot of trust on the developer. If the developer was shady, he would rather hide or try to hide the source code and make it harder for people to find it out.
Since it’s publicly available and used widely enough, there would be ‘those’ people who like finding cracks in code or just have knack for looking deep through all kinds of data.
Not everyone is malicious and that part of humanity is something we have to trust in.
What about the various NPM packages written by one guy. Who then moved on to other things then gave control of that package to someone else that seemed legit. Only for them to slowly add melicious code to that once trusted package that is used by a large number of other packages?
Or someone raising a pull request for a new feature or something that on the surface looks legit on its own. But when combined with other PRs or existing code ends up in a vulnerability that can be exploited.
safe**R** not safe. Seriously how is this a hard concept.
Just that there is ability to read and change the code, even if not everyone reads it, makes developers away from idea to put something malicious there.
Just like how no one has ever put anything malicious on Wikipedia. Nope, never, not once
Wikipedia accepts all new entries by default. Almost all open source projects review any contribiution first before merge.
It’s also not fair comparison, because there can’t exists an encyclopedia you can learn from but not look what’s inside it. But you can obfuscate machine code, making it very hard to see what it does, so it’s more temping for code developers to put malicious features when noone can see it.
And to back by words: https://gnu.org/proprietary/proprietary-back-doors.html
You won’t find such dense lists for anything that has source code available.
This is wrong and ignorant. It happens all the fucking time. Software vendor supply chain is a huge fucking issue.
Christ, tell me you have no idea what your talking about with 1 sentence vibes.
how about you chill? it will happens less frequently than with proprietary software…
Lol no it doesn’t. It happens weekly, all the fucking time.
Source: I’ve been developing oss software for 20 years and have had to push hundreds of teams to fix their vendors bin.
Chill == I ain’t got shit to say 🤣
Get that reddit attitude out of here.
ah, found the redditor.
best of luck im not doing that shit again,you act like a child.
Just an fyi you can block the trolls here.
Hey I know it sucks when someone isn’t nice to you, but that person is about as right as can be.
Just a month ago thousands of malicious commits discovered on git made the news. Unaudited repositories are a huge vector for attack and have been for years.
If that person seems pissed off you could chalk it up to hearing about this stuff on newsgroup discussion two decades ago.
Lololol oSs is free and SeKuR3 cause rainbows and kittens.
20 years of experience and still behaves like a little kid, My 2 year old nephew is more mature. So sad, and ironic that you say that in a foss platform.
With a name like @redditcunts, this one is probably a troll. Just block them.
👌👍
Software vendor supply chain affects ALL software. It is caught much sooner with open source.