The computation of the german wage tax uses religious affiliation as an input. This is necessary to compute the church tax. But does it inadvertedly modify other outputs?
Abstract
This work presents results on the use of Information-Flow tools for the formal verification of algorithmic fairness properties. The problem of enforcing secure information-flow was originally studied in the context of information security: If secret information may “flow” through an algorithm in such a way that it can influence the program’s output, we consider that to be insecure information-flow as attackers could potentially observe (parts of) the secret. Due to its wide-spread use, there exist numerous tools for analyzing secure information-flow properties. Recent work showed that there exists a strong correspondence between secure information-flow and algorithmic fairness: If protected group attributes are treated as secret program inputs, then secure information-flow means that these “secret” attributes cannot influence the result of a program. We demonstrate that off-the-shelf tools for information-flow can be used to formally analyze algorithmic fairness properties including established notions such as (conditional) demographic parity as well as a new quantitative notion named fairness spread.
Samuel Teuber
Doctoral Researcher
Interested in formal methods for software and machine learning verification with a focus on cyber-physical systems and algorithmic fairness.