Gold Doesn't Always Glitter: Spectral Removal of Linear and Nonlinear Guarded Attribute Information
This addresses privacy and fairness concerns in machine learning by enabling more efficient removal of sensitive information, particularly in low-resource scenarios.
The paper tackles the problem of removing private or guarded attribute information from neural representations, introducing a method (Spectral Attribute removaL; SAL) that uses matrix decomposition to project inputs into directions with reduced covariance with the guarded data, and demonstrates it retains better main task performance compared to previous work while requiring relatively small amounts of guarded attribute data.
We describe a simple and effective method (Spectral Attribute removaL; SAL) to remove private or guarded information from neural representations. Our method uses matrix decomposition to project the input representations into directions with reduced covariance with the guarded information rather than maximal covariance as factorization methods normally use. We begin with linear information removal and proceed to generalize our algorithm to the case of nonlinear information removal using kernels. Our experiments demonstrate that our algorithm retains better main task performance after removing the guarded information compared to previous work. In addition, our experiments demonstrate that we need a relatively small amount of guarded attribute data to remove information about these attributes, which lowers the exposure to sensitive data and is more suitable for low-resource scenarios. Code is available at https://github.com/jasonshaoshun/SAL.