ART 1.12.0

[beat-buesser]

This release of ART 1.12.0 introduces the first black-box adversarial patch attack, overlapping shadow datasets for membership inference, certified adversarial training, and more.

Added

• Added Sleeper Agent poisoning attack in TensorFlow in art.attacks.poisoning.SleeperAgentAttack (Implement Sleeper Agent Poisoning Attack in TensorFlow #1769)
• Added support for overlapping shadow models and black-box model predictions as input in membership inference attacks (Shadow models and predictions for membership attack #1778)
• Added adversarial accuracy as a metric (Add adversarial_accuracy as a metric #1779)
• Added function art.utils.uniform_sample_from_sphere_or_ball to sample uniformly from either the ball or the sphere with a given norm and radii (Create sample_from_sphere_or_ball #1804)
• Added GRAPHITE, black- and white-box evasion attacks generating adversarial patches (Implement GRAPHITE attack, fixed #1828)
• Added certified adversarial training (Certified training #1841)

Changed

• Changed art.attacks.evasion.DPatch to accept true labels (DPatch: def parameter updated to accept true labels #1780)
• Changed art.utils.random_sphere to use a different, faster algorithm for norm=1 based on exponential distribution (Update utils.py #1805)

Removed

[None]

Fixed

[None]

---

This discussion was created from the release ART 1.12.0.