Inverting Neural Networks: New Methods to Generate Neural Network Inputs from Prescribed Outputs
This addresses the inverse mapping problem in neural networks, offering insights into model interpretability and security, though it appears incremental as it builds on existing inversion techniques.
The paper tackles the problem of generating neural network inputs that map to specific output classes, introducing two methods that produce random-like images achieving near perfect classification scores, revealing network vulnerabilities.
Neural network systems describe complex mappings that can be very difficult to understand. In this paper, we study the inverse problem of determining the input images that get mapped to specific neural network classes. Ultimately, we expect that these images contain recognizable features that are associated with their corresponding class classifications. We introduce two general methods for solving the inverse problem. In our forward pass method, we develop an inverse method based on a root-finding algorithm and the Jacobian with respect to the input image. In our backward pass method, we iteratively invert each layer, at the top. During the inversion process, we add random vectors sampled from the null-space of each linear layer. We demonstrate our new methods on both transformer architectures and sequential networks based on linear layers. Unlike previous methods, we show that our new methods are able to produce random-like input images that yield near perfect classification scores in all cases, revealing vulnerabilities in the underlying networks. Hence, we conclude that the proposed methods provide a more comprehensive coverage of the input image spaces that solve the inverse mapping problem.