SHLIME: Foiling adversarial attacks fooling SHAP and LIME
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916901187223552 |
|---|---|
| author | Chauhan, Sam Duguet, Estelle Ramakrishnan, Karthik Van Deventer, Hugh Kruger, Jack Subbaraman, Ranjan |
| author_facet | Chauhan, Sam Duguet, Estelle Ramakrishnan, Karthik Van Deventer, Hugh Kruger, Jack Subbaraman, Ranjan |
| contents | Post hoc explanation methods, such as LIME and SHAP, provide interpretable insights into black-box classifiers and are increasingly used to assess model biases and generalizability. However, these methods are vulnerable to adversarial manipulation, potentially concealing harmful biases. Building on the work of Slack et al. (2020), we investigate the susceptibility of LIME and SHAP to biased models and evaluate strategies for improving robustness. We first replicate the original COMPAS experiment to validate prior findings and establish a baseline. We then introduce a modular testing framework enabling systematic evaluation of augmented and ensemble explanation approaches across classifiers of varying performance. Using this framework, we assess multiple LIME/SHAP ensemble configurations on out-of-distribution models, comparing their resistance to bias concealment against the original methods. Our results identify configurations that substantially improve bias detection, highlighting their potential for enhancing transparency in the deployment of high-stakes machine learning systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_11053 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SHLIME: Foiling adversarial attacks fooling SHAP and LIME Chauhan, Sam Duguet, Estelle Ramakrishnan, Karthik Van Deventer, Hugh Kruger, Jack Subbaraman, Ranjan Machine Learning Cryptography and Security Post hoc explanation methods, such as LIME and SHAP, provide interpretable insights into black-box classifiers and are increasingly used to assess model biases and generalizability. However, these methods are vulnerable to adversarial manipulation, potentially concealing harmful biases. Building on the work of Slack et al. (2020), we investigate the susceptibility of LIME and SHAP to biased models and evaluate strategies for improving robustness. We first replicate the original COMPAS experiment to validate prior findings and establish a baseline. We then introduce a modular testing framework enabling systematic evaluation of augmented and ensemble explanation approaches across classifiers of varying performance. Using this framework, we assess multiple LIME/SHAP ensemble configurations on out-of-distribution models, comparing their resistance to bias concealment against the original methods. Our results identify configurations that substantially improve bias detection, highlighting their potential for enhancing transparency in the deployment of high-stakes machine learning systems. |
| title | SHLIME: Foiling adversarial attacks fooling SHAP and LIME |
| topic | Machine Learning Cryptography and Security |
| url | https://arxiv.org/abs/2508.11053 |