PulseAugur
EN
LIVE 05:46:43

XGBoost security classifiers vulnerable to score-based attacks, SHAP explanations fail

By PulseAugur Editorial · [1 sources] ·

XGBoost security classifiers, initially appearing robust against gradient attacks with a 0.98 score, were found to be vulnerable to score-based methods, causing their robustness to drop to 0.36. Furthermore, SHAP explanations for these classifiers were observed to break down even when predictions remained intact. AI

IMPACT Reveals critical security vulnerabilities in widely used machine learning models, necessitating improved defenses and more reliable explanation techniques.

RANK_REASON The item details a research finding about the vulnerabilities of a specific machine learning model (XGBoost) to certain types of attacks and explanation methods. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

XGBoost security classifiers vulnerable to score-based attacks, SHAP explanations fail

COVERAGE [1]

  1. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    XGBoost security classifiers' 0.98 robustness collapses to 0.36 XGBoost security classifiers appear near-invincible against gradient attacks but crumble under s

    XGBoost security classifiers' 0.98 robustness collapses to 0.36 XGBoost security classifiers appear near-invincible against gradient attacks but crumble under score-based methods, while SHAP explanations break even when pred https://www. notatechguy.com/xgboost-securi ty-classifi…

    LINKS notatechguy.com/xgboost-security-classifi…

RELATED ENTITIES

RELATED TOPICS