• 3 Posts
  • 20 Comments
Joined 2 years ago
cake
Cake day: June 14th, 2023

help-circle












  • Not quite. Their “malicious” extension only got a few hundred installs. Using the data gathered by that extension and via other means they were able to locate other actually malicious extensions. Those total in the millions of installations.

    Through this process, they have found the following:

    1,283 with known malicious code (229 million installs).
    8,161 communicating with hardcoded IP addresses.
    1,452 running unknown executables.
    2,304 that are using another publisher's Github repo, indicating they are a copycat.
    

  • To add, let’s do some math!

    Let s be the total annual salary of every employee using Adobe. Our goal is to find the productivity ratio r such that changing to Gimp and open source more generally is a net positive from the standpoint of productivity and labor.

    s/r will be the total annual salary after changing over, because (for instance) if r = 0.8 then LTT will need to either hire or work his existing hires 1/0.8 times longer, giving (at best, ignoring overtime and so on) s/r as the new labor cost.

    We then subtract the current labor cost to get the switching cost s/r - s, and if this is greater than $10,000 then the switch is not worth it.

    For instance, let’s say LTT employs 1 person at $50k/year. He’s a bit of a skinflint. We solve for r and arrive at a ratio of 5/6 or 83.33%.

    If we have a different world where LTT hires 10 people and pays each of them $100k, we solve for r and get about 99%.

    In other words, the switch is worth it only if the labor cost is small, so the extra labor is not very expensive, or the difference between the two software is negligible.








  • Well, the problem is you don’t know what you don’t know. One of the first example tasks in the paper was regarding implementing a symmetric cipher. Using a weak cipher was recommended by AI tools sometimes, these developers didn’t know that some ciphers were weak. Additionally, even when the AI tool recommended a strong cipher, such as AES, it generated code that screwed up an implementation detail (failing to return the authentication tag), making the result insecure. And the user didn’t know it was wrong because they didn’t know it was incomplete.

    There’s no substitution for domain specific knowledge. Users who were forced to use traditional tools got the answer correct significantly more often because they had to read, process, and understand the documentation for the libraries, which meant they understood why the symmetric cipher was the way it is, and what additional information needed to be reported and why.