FBI: AI Makes It Easier for Hackers to Generate Attacks | Hacker News Reader

FBI: AI Makes It Easier for Hackers to Generate Attacks

tomshardware.com

14 pointsthunderbong11 comments

AI makes it easier for anyone to write software. So, yeah that makes sense, I don't think it warrants an article, and I think the FBI might just be fear mongering.

A list of things that make is easier for hackers to generate attacks:

* Vim

* GCC

* Google

* Intel's Documentation

* Math class

None of your list are recent changes. Changes to the ease, and therefore likely incidence/severity, of certain attacks are fairly normal things for the FBI to be concerned about.

I believe they should have instead phrased it something along the line of "AI facilitates a new generation of Skiddies [1] lowering the bar even further to malicious behavior without much technical knowledge." or something along that line. They're not wrong.

Now a person with little or no technical knowledge can ask an LLM in plain English how to generate a more complex script that does {x+y+z} vs. at least having to know some degree of a scripting language or being limited to one of the well known and defended against skiddie tools that only perform {x}. A theoretical example would be a low-tech person taking full control of an internet connected car on a live highway. It's been done by technically advanced researchers [2] but now the bar to entry is now a lot lower and LLM's may likely keep up with the latest working exploits. That specific exploit was patched but there will always be new ones.

[1] - https://en.wikipedia.org/wiki/Skiddies

[2] - https://www.youtube.com/watch?v=OobLb1McxnI [video][46 mins]

On the other hand, GPTs often produce buggy code, and script kiddies are a lot less likely to be able to debug AI-generated code than they are to be successful running another human's hand-written exploit.

I could envision LLM's taking that even further given I am sure they will be used to manipulate people much in the same manor as social media algorithms, I could see skiddies being detected and then tricked into running code that will either not work or will brick their own devices assuming they clicked their way through a waiver of liability of damages at some point.

It allows for even more targeted phishing/scams and lowers the bar of entry required to do this and the difficulty of doing this at scale. the human element is a pretty common entry point.

This argument is like saying that Napster didn't change anything because we'd previously been able to trade MP3s on FTP servers.

How do you feel about the 2nd amendment and bans on automatic weapons? /s

Does AI make it easier to patch security holes and to identify attacks?

Why only focus on one side?

I for one am excited to truly have intelligent "ICE" like from Gibson's Neuromancer (or in its more modern incarnation like in Cyberpunk 2077).

gov scared of something that gives its population more power better work to undermine it!!!

Reply on news.ycombinator.com