Trustworthy AI by Beena Ammanath

Author:Beena Ammanath [Ammanath, Beena]
Language: eng
Format: epub
ISBN: 9781119867951
Publisher: Wiley
Published: 2022-03-07T00:00:00+00:00

What Does AI Compromise Look Like?

Many business leaders understand the core elements in cybersecurity: strong access credentials, active system monitoring, and employee training for effective cyber hygiene and avoiding social engineering attacks. Those elements are insufficient for AI security due to how AI systems operate and evolve, as well as the novel attack vectors cybercriminals and bad actors are still discovering.

These are the early days of AI security. Much focus is put on basic taxonomy, classifying how security may be circumvented, and from there, deciphering how it can be mitigated. As a starting point, consider a three‐axis taxonomy as conceived by researchers at the University of California, Berkeley:5

Influence – An attack can be causative (i.e., influencing the data) or exploratory (i.e., observing how AI responds to an action).

Security violation – The attack can yield false negatives (integrity violation) or false positives (availability violation).

Specificity – The attack can be focused on a specific classifier in an AI system (targeted) or directed to classifiers across a system or many systems (indiscriminate).

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.