Building our threat library
Before delving into threat modeling, we must define our AI threats dictionary and relate it to Adversarial AI. So far, we have used the terms threat and attack interchangeably. Let’s remind ourselves of what each of them means:
- Threats: A threat is a potential cause of harm to a system through intentional actions such as hacking or accidental occurrences such as natural disasters. It signifies possible danger but not the action itself.
- Attacks: An attack involves actively exploiting vulnerabilities in a system, aiming to breach security, inflict harm, or steal data. It’s the practical execution of a threat.
- Vulnerabilities: Vulnerabilities are flaws or weaknesses in a system that can be exploited to gain unauthorized access or cause damage. They are the specific openings that are targeted in attacks.
- Risks: Risk measures the potential for loss or damage when a threat exploits a vulnerability, combining the likelihood of...
