Audit Algorithms
91 items from erwanlemerrer/awesome-audit-algorithms ★122
-
TamperNN: Efficient Tampering Detection of Deployed Neural Nets arxiv.org
(ISSRE) Algorithms to craft inputs that can detect the tampering with a remotely executed classifier model.
-
Knockoff Nets: Stealing Functionality of Black-Box Models arxiv.org
(CVPR) Ask to what extent can an adversary steal functionality of such "victim" models based solely on blackbox interactions: image in, predictions out.
-
Copycat CNN: Stealing Knowledge by Persuading Confession with Random Non-Labeled Data arxiv.org
(IJCNN) (Code) Stealing black-box models (CNNs) knowledge by querying them with random natural images (ImageNet and Microsoft-COCO).
-
Adversarial Frontier Stitching for Remote Neural Network Watermarking arxiv.org
(Neural Computing and Applications) (Alternative implementation) Check if a remote machine learning model is a "leaked" one: through standard API requests to a remote model, extract (or not) a zero-bit watermark, that was inserted to watermark valuable models (eg, large deep…
-
Towards Reverse-Engineering Black-Box Neural Networks arxiv.org
(ICLR) (Code) Infer inner hyperparameters (eg number of layers, non-linear activation type) of a remote neural network model by analysing its response patterns to certain inputs.
-
Counterfactual Explanations without Opening the Black Box: Automated Decisions and the GDPR arxiv.org
(Harvard Journal of Law & Technology) To explain a decision on x, find a conterfactual: the closest point to x that changes the decision.
-
Distill-and-Compare: Auditing Black-Box Models Using Transparent Model Distillation arxiv.org
(AIES) Treats black box models as teachers, training transparent student models to mimic the risk scores assigned by black-box models.
-
The topological face of recommendation: models and application to bias detection arxiv.org
(Complex Networks) Proposes a bias detection framework for items recommended to users.
-
Auditing Black-Box Models for Indirect Influence arxiv.org
(ICDM) Evaluate the influence of a variable on a black-box model by "cleverly" removing it from the dataset and looking at the accuracy gap
-
Certifying and Removing Disparate Impact arxiv.org
(SIGKDD) Proposes SVM-based methods to certify absence of bias and methods to remove biases from a dataset.
-
Algorithmic Transparency via Quantitative Input Influence: Theory and Experiments with Learning Systems andrew.cmu.edu
(IEEE S&P) Evaluate the individual, joint and marginal influence of features on a model using shapley values.
