Last week, I wrote about catching a supply chain attack on a WordPress plugin called Widget Logic. A trusted name, acquired by a new owner, turned into
The WordPress plugin marketplace has a trust problem.
I think that the problem is really broader — that for any system, be it Linux distros or browser plugins or AI Python packages or NPM packages or whatever — even trustworthy software can change ownership. Most users are probably not monitoring those changes and are not in a position to evaluate the impact of those changes.
Some of that can (and probably should) be handled by compartmentalizing software, limiting the effect it can have, though that has some costs of its own. But I don’t think that that’s going to handle everything.
Yeah and “change ownership” can be subtle, like a compromised account, so there’s no changes to monitor apart from an update (and you expect updates regularly anyway).
I do think the linux philosophy of “do one thing and do it well” can help a lot, especially when paired with restrictive permissions. But when everything and their dog is integrating LLMs, that alone is already becoming harder.
In open source you could inspect every update (and probably some people do), but unless you’re an above average programmer you probably won’t understand everything, let alone catch hidden backdoors. And then there’re supply chain attacks that don’t change the source code but do change the compiled version so good luck checking that out manually each time unless you always build from scratch.
I think that the problem is really broader — that for any system, be it Linux distros or browser plugins or AI Python packages or NPM packages or whatever — even trustworthy software can change ownership. Most users are probably not monitoring those changes and are not in a position to evaluate the impact of those changes.
Some of that can (and probably should) be handled by compartmentalizing software, limiting the effect it can have, though that has some costs of its own. But I don’t think that that’s going to handle everything.
Yeah and “change ownership” can be subtle, like a compromised account, so there’s no changes to monitor apart from an update (and you expect updates regularly anyway).
I do think the linux philosophy of “do one thing and do it well” can help a lot, especially when paired with restrictive permissions. But when everything and their dog is integrating LLMs, that alone is already becoming harder.
In open source you could inspect every update (and probably some people do), but unless you’re an above average programmer you probably won’t understand everything, let alone catch hidden backdoors. And then there’re supply chain attacks that don’t change the source code but do change the compiled version so good luck checking that out manually each time unless you always build from scratch.