Hybrid reverse knowledge distillation for adversarial example detection

Hyun Kwon, Joo Bon Maeng, Dae-Jin Kim · 2026

Read Hybrid reverse knowledge distillation for adversarial example detection by Hyun Kwon, Joo Bon Maeng, Dae-Jin Kim (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13328724.txt.

Description

Hyun Kwon, Joo Bon Maeng, and Dae-Jin Kim present a two thousand twenty-six method for detecting adversarial examples crafted to fool image classifiers. Tiny invisible tweaks flip model predictions, endangering medical and driving systems. Detectors must catch inputs that look normal to humans.

Their system trains a decoder to rebuild a frozen network's internal features from clean images only. Attack inputs distort those features, so reconstruction fails measurably. Statistical distance checks add a second alarm. Tests on photo and skin-image datasets against five attack families reached strong detection scores, advancing practical AI safety screening.

Catalog record

CREATORHyun Kwon, Joo Bon Maeng, Dae-Jin Kim
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC13328724

Browse the full library · About the Tower of Babel