219
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 20 Jun 2023
219 points (100.0% liked)
Technology
37728 readers
612 users here now
A nice place to discuss rumors, happenings, innovations, and challenges in the technology sphere. We also welcome discussions on the intersections of technology and society. If it’s technological news or discussion of technology, it probably belongs here.
Remember the overriding ethos on Beehaw: Be(e) Nice. Each user you encounter here is a person, and should be treated with kindness (even if they’re wrong, or use a Linux distro you don’t like). Personal attacks will not be tolerated.
Subcommunities on Beehaw:
This community's icon was made by Aaron Schneider, under the CC-BY-NC-SA 4.0 license.
founded 2 years ago
MODERATORS
Hunh.
I just had a surge of user registrations on my instance.
All passed the captcha. All passed the email validation.
All, had a valid-sounding response.
I am curious to know if they are actual users, or.... if I just became the host of a spam instance. :-/
Doesn't appear to be an easy way to determine.
Hmmm, I'd check the following:
With those answers I should be able to tell if it's the same or similar attacker getting more sophisticated.
Some patterns I noticed in the attacks I've received:
Some vulnerabilities I know that can be exploited and would expect to see next:
I think it would be interesting if we could find a prompt that doesn't work well with LLMs. Originally they struggled with math for example, but I wonder if it'd be possible to make a math problem that's simple enough for most humans to solve but which trips up LLMs into outputting garbage.
I personally use this to track who send my email address to where, since people usually don't strip this from the address. It's definitely abusable, but also has legitimate uses.
Not so sure on the LLM front, GPT4+Wolfram+Bing plugins seems to be a doozy of a combo. If anything there should be perhaps a couple interactable elements on the screen that need to be interacted with in a dynamic order that's newly generated for each signup. Like perhaps "Select the bubble closest to the bottom of the page before clicking submit" on one signup and "Check the box that's the furthest to the right before clicking submit"?
Just spitballin it there.
As for the category on email address - certainly not suggesting they remove supporting it, buuuuutttt if we're all about making sure 1 user = 1 email address, then perhaps we should make the duplication check a bit more robust to account for these types of emails. After all someuser+lemmy@somedomain.com is the same as someuser@somedomain.com but the validation doesn't see that. Maybe it should?
I like your idea of interaction-based authentication. Extra care would need to go into making sure it's accessible, but otherwise I think that would be a stronger challenge for LLMs to solve. (Keep in mind LLMs can still receive the page's HTML as context, but that seems like it could present as a stronger challenge even still.)
This makes sense to me. I could be wrong, but the assumption of 1 email = 1 user doesn't seem unreasonable, especially since there's no cost to making a new email address.