Penetration Testing the Artificial Intelligence

Artificial Intelligence

Artificial intelligence algorithms can be infiltrated and controlled by an adversary.

The systems underpinning the state-of-the-art artificial intelligence systems are systematically vulnerable to a new type of adversarial attack called an “artificial intelligence attack” or also called Adversarial Machine Learning.

Aritificial Intelligence (AI)
Machine Learning (ML)
Deep Learning
Neural Learning
Espionage
Sabotage
fraud
Inference by Poisoning Poisoning Poisoning
Inference Attacks Trojaning Evasion
Backdooring False Positives Evasion

Types of Adversarial Attacks on AI

  1. Espionage.
  2. Sabotage.
  3. Machine Learning Fraud.
penetration testing methodologies
Pen Testing the AI Models
AI Poisoning Attacks
AI Backdooring
Trojans on AI Models
DDoSing AI with Requests
Attacks on AI Supervised Learning
Parameter Inference – ML/AI Model Extraction