Auditing AI models for Verified Deployment under Semantic Specifications
arXiv:2109.12456
Abstract
Auditing trained deep learning (DL) models prior to deployment is vital for preventing unintended consequences. One of the biggest challenges in auditing is the lack of human-interpretable specifications for the DL models that are directly useful to the auditor. We address this challenge through a sequence of semantically-aligned unit tests, where each unit test verifies whether a predefined specification (e.g., accuracy over 95%) is satisfied with respect to controlled and semantically aligned variations in the input space (e.g., in face recognition, the angle relative to the camera). We enable such unit tests through variations in a semantically-interpretable latent space of a generative model. Further, we conduct certified training for the DL model through a shared latent space representation with the generative model. With evaluations on four different datasets, covering images of chest X-rays, human faces, ImageNet classes, and towers, we show how AuditAI allows us to obtain controlled variations for certified training. Thus, our framework, AuditAI, bridges the gap between semantically-aligned formal verification and scalability. A blog post accompanying the paper is at this link https://developer.nvidia.com/blog/nvidia-research-auditing-ai-models-for-verified-deployment-under-semantic-specifications
arXiv commentPreprint; Under review
References in corpus (26)
- Learning Transferable Visual Models From Natural Language Supervision
- Large Scale GAN Training for High Fidelity Natural Image Synthesis
- Language Models are Few-Shot Learners
- Progressive Growing of GANs for Improved Quality, Stability, and Variation
- LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop
- CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning
- DenseNet: Implementing Efficient ConvNet Descriptor Pyramids
- Certified Adversarial Robustness via Randomized Smoothing
- On Evaluating Adversarial Robustness
- Generating Natural Adversarial Examples
- Understanding the Role of Individual Units in a Deep Neural Network
- PixelDefend: Leveraging Generative Models to Understand and Defend against Adversarial Examples
- Towards a Definition of Disentangled Representations
- On the Effectiveness of Interval Bound Propagation for Training Verifiably Robust Models
- Defense-GAN: Protecting Classifiers Against Adversarial Attacks Using Generative Models
- Spatially Transformed Adversarial Examples
- Generating Adversarial Examples with Adversarial Networks
- Semidefinite relaxations for certifying robustness to adversarial examples
- Efficient Formal Safety Analysis of Neural Networks
- Smooth Adversarial Training
- Closed-Form Factorization of Latent Semantics in GANs
- Spatially-Aware Graph Neural Networks for Relational Behavior Forecasting from Sensor Data
- Characterizing Adversarial Examples Based on Spatial Consistency Information for Semantic Segmentation
- Learning perturbation sets for robust machine learning
- Algorithms for Verifying Deep Neural Networks
- Certified Robustness to Adversarial Word Substitutions