BDFProxy VS quicklisp-client

for same sort of thing but not lisp, see backdoor factory that will backdoor any .exe you download over connection that attacker is MITMing. attacker doesn't need to know what specific library you will download from quicklisp, they write a mitmproxy script like that so for any download from quicklisp.org, it opens the .tar.gz, adds some malicious lisp to it (probably that just executes shell command to download and execute their normal malware, password stealer or whatever as I don't think they going to write full malware in lisp), repack it as .tar.gz you were requesting and serve it to you. It's not the same issue as phishing where they email saying please open and run attachment.exe and you click through all the warnings that you are doing something dangerous and about to run untrusted code. You just use quicklisp as you normally do, if you install any package, when an attacker can MITM your connection they can run code on your computer. Yes that is sometimes also possible with browser exploit but browsers have multiple layers of sandbox and protections against it, and when someone finds a vulnerability that gets through it is treated as a serious vulnerability to fix. some of this thread seems people saying well nothing is perfectly secure a sufficiently pacient, skilled, well-funded attacker can always get through somehow, so it doesn't matter raising the bar off the floor by not using http unverified to download code we run on people's computer

Have you seen this mitmproxy plugin: https://github.com/secretsquirrel/BDFProxy They use mitmproxy to capture and replace software auto-updates.

Yes, that's clear.

I'm not very familiar with how quicklisp works. I thought that “updates once a month” implies a separate update channel (distribution, ...).

Looking at the relevant issue, https://github.com/quicklisp/quicklisp-client/issues/167 , it's not clear that even hashes are in place.

I recently found out that most Nix fetchers use https, but do not actually do verification (`curl --insecure` or equivalent libcurl settings). Channel updates do verify and include hashes, so the overall chain is authenticated.

The latest comment I see about this here from Oct. 2022 says they're working on it. There's also comment by the developer in 2016 saying want to improve the security soon, so it doesn't really seem this will actually happen soon. I realise making signature verification work cross platform in pure lisp without external dependencies isn't easy but from latest comment it seems they have that working, in a branch written 4 years ago? The simplest no-code solution is just since quicklisp is published every month or so, on each new update publish a file with sha256 hash of every package contained in quicklisp signed with same developer's pgp key they are already using to sign download of the initial quicklisp.lisp, yes then users if they care about security would have to manually download the file and verify signature every month or so but it's at least some solution that can be done now.

> That's what regular devs do, they don't even bother writing articles or commenting on HN :-)

I'll take the bait, and roll up several of my comments into one.

First, the support contract costs from the commercial vendors can make sense. It's one of the most expensive parts of software. We joke about fixing relatives' printers, but its not false. Support costs introduce a counter-balance.

Second, a message to everyone looking into or using QuickLisp, it uses http instead of https: https://github.com/quicklisp/quicklisp-client/issues/167

You can patch your version to fix this. I'd also recommend adding firewall rules to deny in case your patches roll back. And any other mitigation. Or stricter policies, such as not using it, if it makes sense for your organization.

And the AI bots? I hope there aren't people herding them who don't want to, that's how you get unloving brats and a crappy world.

I found this github issue about it, open since 2018: https://github.com/quicklisp/quicklisp-client/issues/167

I agree 100% about needing to test and audit for security. But based on the information I've seen and public activity in repos, I assumed Xach was going for home-grown CL implementation. https://github.com/quicklisp/quicklisp-client/blob/pgp/quicklisp/openpgp.lisp