Presentation 2022-03-09
Fairness Testing of Machine Learning Software through a Combinatorial Approach
Daniel Perez Morales, Takashi Kitamura, Shingo Takada,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Machine learning (ML) can be used in decision-making algorithms or classifiers. These classifiers must be tested looking for discrimination. It is an unintended behavior and can disfavor certain individuals based on their protected attributes, such as race or gender. Aequitas, a well-known black-box technique for fairness testing, tests ML software to find discrimination. Although Aequitas has several technical advantages, it largely relies on random sampling. We propose Coverage-Guided Fairness Testing (CGFT). CGFT applies combinatorial t-way testing when searching for discriminatory data. We evaluate CGFT with Aequitas, showing an improvement in the number of discrimination found using CGFT.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Fairness testing / Combinatorial testing / Machine learning
Paper # KBSE2021-50
Date of Issue 2022-03-02 (KBSE)

Conference Information
Committee KBSE
Conference Date 2022/3/9(2days)
Place (in Japanese) (See Japanese page)
Place (in English) Online
Topics (in Japanese) (See Japanese page)
Topics (in English) General, Student
Chair Hiroyuki Nakagawa(Osaka Univ.)
Vice Chair Takuya Saruwatari(NTT Data)
Secretary Takuya Saruwatari(Shinshu Univ)
Assistant Hideharu Kojima(Osaka Univ.) / Yutaro Kashiwa(Kyushu Univ,)

Paper Information
Registration To Technical Committee on Knowledge-Based Software Engineering
Language ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Fairness Testing of Machine Learning Software through a Combinatorial Approach
Sub Title (in English)
Keyword(1) Fairness testing
Keyword(2) Combinatorial testing
Keyword(3) Machine learning
1st Author's Name Daniel Perez Morales
1st Author's Affiliation National Institute of Advanced Industrial Science and Keio Univ.(AIST and Keio Univ.)
2nd Author's Name Takashi Kitamura
2nd Author's Affiliation National Institute of Advanced Industrial Science(AIST)
3rd Author's Name Shingo Takada
3rd Author's Affiliation Keio University(Keio Univ.)
Date 2022-03-09
Paper # KBSE2021-50
Volume (vol) vol.121
Number (no) KBSE-424
Page pp.pp.54-59(KBSE),
#Pages 6
Date of Issue 2022-03-02 (KBSE)