NIST and Adversarial Machine Learning | Signal and the art of communication
#179 - Predictive and Generative AI Taxonomy | Using Signal is secure, till you add journalists to your group on war plans
NIST releases a clear taxonomy for adversarial machine learning terminologies
If you want to know the difference between ‘model poisoning’ and ‘data poisoning’ in AI, this is the document to read.
This document by NIST defines the taxonomy (definitions) of different terms in adversarial machine learning.
For example, both predictive AI and generative AI have “Supply Chain Attacks” defined. The difference between the two lies in the impacts that the attack would have.
“Misuse violation” is possible for generative AI. Predictive AI has limited risk of being misused.
Take Action:
If you are responsible for conducting any sort of risk assessments on AI systems, this is a document that you must know. Not only does it define the terminologies, it also clarifies the structure of possible compromises to AI systems.
First, close the doors; How to use Signal 101 🤣
Using a secure encrypted algorithm does not guarantee security, just like having an pick proof lock does not guarantee security if you just leave the door open
A journalist was inadvertently added to a Signal group containing top US government officials who were planning a military strike!! It’s hilarious! (the inadvertent addition, not military strikes) Just read it yourself.
Then, the NSA got in and said that they had released a document about the vulnerabilities of Signal a while back.
Then Signal said that the NSA document is a tad inappropriately titled, because while it says ‘vulnerabilities of signal’ it just talks about social engineering and stuff.
Take Action:
Learn from your teenager 🤣:
Check the members of any group before sending sensitive messages
Do not add people you do not know to groups
Don’t use publicly available chat software for highly confidential government stuff (duh!)
If you inadvertently add someone and they get hold of sensitive information, do not blame them for being a part of it.
The 🔮 Oracle Cloud story is still developing. I will follow it and update you folks.