paper-with-me

Papers

Towards Reverse-Engineering Black-Box Neural Networks

2017-11-06 · ICLR 2018 1 · Seong Joon Oh, Max Augustin, Bernt Schiele, Mario Fritz

Many deployed learned models are black boxes: given input, returns output. Internal information about the model, such as the architecture, optimisation procedure, or training data, is not disclosed explicitly as it might contain proprietary information or make the system more vulnerable. This work shows that such attributes of neural networks can be exposed from a sequence of queries. This has multiple implications. On the one hand, our work exposes the vulnerability of black-box neural networks to different types of attacks -- we show that the revealed internal information helps generate more effective adversarial examples against the black box model. On the other hand, this technique can be used for better protection of private content from automatic recognition models using adversarial examples. Our paper suggests that it is actually hard to draw a line between white box and black box models.

📄 PDF Abstract BibTeX arXiv:1711.01768

Code (3)

KuoTzu-yang/ML-reverse pytorch
coallaoh/whitenblackbox pytorch
markliou/model_distillation tf

Similar Papers 제목 키워드 기반

DREAM: Domain-agnostic Reverse Engineering Attributes of Black-box Model

2024-12-08 · Rongqing Li, Jiaqi Yu, Changsheng Li, Wenhan Luo 외

Deep learning models are usually black boxes when deployed on machine learning platforms. Prior works have shown that the attributes (e.g., the number of convolutional layers) of a target black-box model can be exposed t…

Attribute

DREAM: Domain-free Reverse Engineering Attributes of Black-box Model

2023-07-20 · Rongqing Li, Jiaqi Yu, Changsheng Li, Wenhan Luo 외

Deep learning models are usually black boxes when deployed on machine learning platforms. Prior works have shown that the attributes ($e.g.$, the number of convolutional layers) of a target black-box neural network can b…

Attribute

IReEn: Reverse-Engineering of Black-Box Functions via Iterative Neural Program Synthesis

2020-06-18 · NeurIPS Workshop CAP 2020 12 · Hossein Hajipour, Mateusz Malinowski, Mario Fritz

In this work, we investigate the problem of revealing the functionality of a black-box agent. Notably, we are interested in the interpretable and formal description of the behavior of such an agent. Ideally, this descrip…

Computer SecurityProgram Synthesis

Black-Box Guardrail Reverse-engineering Attack

2025-11-06 · Hongwei Yao, Yun Xia, Shuo Shao, Haoran Shi 외 arxiv

Large language models (LLMs) increasingly employ guardrails to enforce ethical, legal, and application-specific constraints on their outputs. While effective at mitigating harmful responses, these guardrails introduce a …

Reinforcement LearningData Augmentation

Reverse Engineering Deep ReLU Networks An Optimization-based Algorithm

2023-12-07 · Mehrab Hamidi

Reverse engineering deep ReLU networks is a critical problem in understanding the complex behavior and interpretability of neural networks. In this research, we present a novel method for reconstructing deep ReLU network…

L2 Regularization