ipfs sounds really interesting. haven't read much about it - what prevents a malicious node from attacking users?

like someone's node saying they have the website you're looking for, but they've added additional malicious code.

Content-addressing is what makes this impossible.

The attack you describe is possible with HTTP, but impossible with IPFS.

HTTP retrieves files based on where they are. Switch the files at that location, and the user downloads malicious files instead.

IPFS retrieves files based on what they are. So you look for a file hash, and it retrieves the exact file you ask for, regardless of where it is.