ART 1.10.0

[beat-buesser]

This release of ART 1.10.0 introduces multiple poisoning attacks on image classification and deep generative models, the first attack with dynamic patches on object tracking in videos, classification certification based on zonotope representations, EoT support for object detection in image rotation and center cropping, new features for attribute inference attacks and more.

Added

• Added Gradient Matching (Witches' Brew) attack art.attacks.poisoning.GradientMatchingAttack in TensorFlow (Gradient matching (Witches' Brew) for Tensorflow #1587)
• Added functions projection_l1_1 and projection_l1_2 to art.utils for two algorithms computing orthogonal projections on L1-norm balls (Add functions for orthogonal projection on L1 balls #1586)
• Added perspective transformations to art.attacks.evasion.AdversarialTexturePyTorch attack to enable dynamic texture/patches (Add support for perspective transformation of texture in AdversarialTexturePyTorch #1557)
• Added support for object detection in art.attacks.evasion.AdversarialPatchPyTorch (Add support for object detection in AdversarialPatchPyTorch #1535)
• Added new features to attribute inference attacks including support for optional use of true labels in black-box attribute inference attacks, automatic calculation of values in fit() method, additional scaling method for labels/predictions and an additional attack model type (random forest) (Improvements to attribute inference attacks #1534)
• Added estimator art.estimators.certification.PytorchDeepZ based on DeepZ for robustness certification using zonotope representations datapoints (DeepZ Certification #1531)
• Added Expectation over Transformation (EoT) for rotation and centre crop with support for classification and object detection (Add EoT center crop and rotation for object detection in PyTorch #1516)
• Added support for SummaryWriter in art.attacks.evasion.RobustDpatch (Add support for Summary Writer in RobustDPatch #1513)
• Added PGD L-Inf optimizer to art.attacks.evasion.AdversarialPatch* attacks (Add PGD-L_Inf optimizer to AdversarialPatch* attacks #1495)
• Added two backdoor poisoning attacks, Red in art.attacks.poisoning.BackdoorAttackDGMReD and Trail in art.attacks.poisoning.BackdoorAttackDGMTrail, targeting Deep Generative Models (Implement backdoor attacks on DGMs #1490)
• Added Hidden Trigger Backdoor Poisoning Attack in Keras and PyTorch in art.attacks.poisoning.HiddenTriggerBackdoor (Hidden trigger Backdoor Poisoning Attack Implementation #1487)
• Added Feature Collision Poisoning Attack in PyTorch in `art.attacks.poisoning.FeatureCollisionAttack (Implemented Poisoning Attack - Poison Frog in PyTorch Framework #1435 )

Changed

• Changed imports of TensorFlow v2 in TensorFlowClassifier to support TensorFlow v1 compatibility mode (Extend support for TensorFlow v1 compatibility mode #1560)
• Changed Python used for unit testing to newer versions, upgraded style checks and improved code quality (Update Python and dependencies #1517)

Removed

[None]

Fixed

• Fixed import of Scipy in PixelThreshold attack to support scipy>=1.8 (Check scipy version for PixelThreshold attack #1589)
• Fixed bug of missing attribute in PixelAttack for scaled images (fix for rescale not set bug #1574)
• Fixed use of torchaudio.functional.magphase in PyTorchDeepSpeech to support Deep Speech 2 version 3 with torch>=1.10 (Fixes torchaudio.functional.magphase use for Deep Speech 2 version 3 #1550)
• Fixed method fitof ScikitlearnRegressor to process labels correctly (Fix ScikitlearnRegressor fit method #1537)
• Fixed scalar names of Indicators of Attack Failure 2 and 3 for aggregated losses (Fix scalar names for indicators 2 and 3 for aggregated losses #1512)
• Fixed raising of DataConversionWarning in fitting black box membership inference attacks with attack_model_type 'rf' or 'gb (Fix fitting warning when using attack_model_type 'rf' or 'gb' #1488)

---

This discussion was created from the release ART 1.10.0.