AI Security Evaluator | AI Governance Online

AI Security Evaluator

This project is to test and validate how resilient the given AI model is against attack algorithms.

Choose a neural network model and an adversarial attack algorithm, Click the Submit button, and receive a report on the attack test results.

Start Test

To start your test, please select a neural network model, then select an adversarial attach algorithm.

Select neural network model to test:

Select testing dataset size:

Select sample: If no sample is selected, the system will randomly choose one as the sample.

Select sample: If no sample is selected, the system will randomly choose one as the sample.

Using targeted attack:

Choose an adversarial attack algorithm to attack the neural network model:

Partner Project:

ARES (Adversarial Robustness Evaluation for Safety)