Posts

AI Attack

Phishing, ransomware, DDoS, Viruses, and Attack vectors are the most common cyber attack type(1). With the expansion of using AI in It’s predictable, we see more and more AI attacks shortly. 

These “AI attacks” are fundamentally different from traditional cyberattacks(2), and current approaches are using against current cyberattacks that may not be appliable. AI Attacks may follow below goals:

  • Cause Damage: the attacker wants to cause damage by having the AI system malfunction. An example of this is an attack to cause an autonomous vehicle to ignore stop signs.
  • Hide Something: the attacker wants to evade detection by an AI system.
  • Degrade Faith in a System: the attacker wants an operator to lose faith in the AI system

Currently, The most common method used in AI Attacks is input attack, which makes a compromised AI system. It refers to manipulating what is fed into the AI system to alter the system’s output to serve the attacker’s goal(2).

We may see the following products in the near future:

  • Data Anomaly Detector made by computer security companies
  • AI training security certificate that ensures the quality of the AI System. (Five Star AI System ).

 

References

1.https://www.itgovernance.co.uk/blog/different-types-of-cyber-attacks

2. https://www.belfercenter.org/publication/AttackingAI

 

Input Attack to make Compromised AI