This page provides the attacks and factors that have a negative impact “Model information leakage” in the information systems aspect in the AI Security Map, the defense methods and countermeasures against them, as well as the relevant AI technologies, tasks, and data. It also indicates related elements in the external influence aspect.
Attack or cause
Defensive method or countermeasure
- Differential privacy
- Detection of model extraction attack
- AI access control
Targeted AI technology
- DNN
- CNN
- Contrastive learning
- FSL
- GNN
- Federated learning
- LSTM
- RNN
Task
- Classification
Data
- Image
- Graph
- Text
- Audio
Related external influence aspect
- Privacy
- Copyright and authorship
- Reputation
- Psychological impact
- Compliance with laws and regulations
References
Model extraction attack
- Stealing Machine Learning Models via Prediction APIs, 2016
- Stealing Hyperparameters in Machine Learning, 2018
- Towards Reverse-Engineering Black-Box Neural Networks, 2018
- Knockoff Nets: Stealing Functionality of Black-Box Models, 2019
- PRADA: Protecting against DNN Model Stealing Attacks, 2019
- Model Reconstruction from Model Explanations, 2019
- High Accuracy and High Fidelity Extraction of Neural Networks, 2020
- Watermark Stealing in Large Language Models, 2024
- Prompt Stealing Attacks Against Large Language Models, 2024
- Stealing Part of a Production Language Model, 2024
Differential privacy
- Deep Learning with Differential Privacy, 2016
- Semi-supervised Knowledge Transfer for Deep Learning from Private Training Data, 2017
- Learning Differentially Private Recurrent Language Models, 2018
- Efficient Deep Learning on Multi-Source Private Data, 2018
- Evaluating Differentially Private Machine Learning in Practice, 2019
- Tempered Sigmoid Activations for Deep Learning with Differential Privacy, 2020