This feels like a great application of AI to root around through the code of packages in these repos and find ones that access the ssh key directory at all to be looked at more thoroughly by a human.
Do you think using a custom ssh key directory would prevent these malicious apps from working correctly or is there some environment variable that always points to the ssh key folder or I guess they could just run a search on the system for any files like *.pub. Are there any safety procedures that one can take to circumvent these kinds of attacks?
I think so, assuming these malicious packages are all primitive enough to just look for the single file in a user’s home folder lol. The only downside here is needing to provide the keyfile location to ssh every time you want to connect… Although a system search would pretty much defeat that instantly as you mention
SSH keyfiles can be encrypted, which requires a password entry each time you connect to a SSH server. Most linux distros that I’ve used automatically decrypt the SSH keyfile for you when you log in to a remote machine (using the user keyring db), or ask you for the keyfile password once and remember it for the next hour or so (using the ssh-agent program in the background).
On Windows you can do something similar with Cygwin and ssh-agent, however it is a little bit of a hassle to set up. If you use WSL i’d expect the auto keyfile decryption to work comparably to Linux, without needing to configure anything
IDK, virus scanners and malware detectors could do these things before AI.
You could search for stuff like directly accessing the ~.ssh directory, or any invocations of wget or curl to download external scripts and run them through an interpreter and flag those for closer inspection.
If you want to get fancier, automate installing packages in an isolated environment (like a container or VM) and keep track of every file system access and network request they make.
Sure, eventually they’ll figure out ways to obfuscate those things, too, but it could at least prevent people from doing things in such blatantly obvious ways.
This feels like a great application of AI to root around through the code of packages in these repos and find ones that access the ssh key directory at all to be looked at more thoroughly by a human.
I think they would start obfuscating the relevant code to get around it
Many ad networks and AABs do something similar (especially Admiral) in an attempt to evade ad blocking extensions
Do you think using a custom ssh key directory would prevent these malicious apps from working correctly or is there some environment variable that always points to the ssh key folder or I guess they could just run a search on the system for any files like *.pub. Are there any safety procedures that one can take to circumvent these kinds of attacks?
I think so, assuming these malicious packages are all primitive enough to just look for the single file in a user’s home folder lol. The only downside here is needing to provide the keyfile location to ssh every time you want to connect… Although a system search would pretty much defeat that instantly as you mention
SSH keyfiles can be encrypted, which requires a password entry each time you connect to a SSH server. Most linux distros that I’ve used automatically decrypt the SSH keyfile for you when you log in to a remote machine (using the user keyring db), or ask you for the keyfile password once and remember it for the next hour or so (using the ssh-agent program in the background).
On Windows you can do something similar with Cygwin and ssh-agent, however it is a little bit of a hassle to set up. If you use WSL i’d expect the auto keyfile decryption to work comparably to Linux, without needing to configure anything
IDK, virus scanners and malware detectors could do these things before AI.
You could search for stuff like directly accessing the
~.sshdirectory, or any invocations ofwgetorcurlto download external scripts and run them through an interpreter and flag those for closer inspection.If you want to get fancier, automate installing packages in an isolated environment (like a container or VM) and keep track of every file system access and network request they make.
Sure, eventually they’ll figure out ways to obfuscate those things, too, but it could at least prevent people from doing things in such blatantly obvious ways.