Explainable Label-flipping Attacks on Human Emotion Assessment System

TL;DR

This paper investigates label-flipping data poisoning attacks on EEG-based human emotion classification systems, demonstrating attack success across models and using XAI techniques to explain these vulnerabilities.

Contribution

It introduces two label-flipping attack scenarios on EEG emotion classifiers and applies XAI methods to elucidate attack impacts.

Findings

Attacks are effective across different classifiers.

Models exhibit varying resistance levels.

XAI techniques help explain attack effects.

Abstract

This paper's main goal is to provide an attacker's point of view on data poisoning assaults that use label-flipping during the training phase of systems that use electroencephalogram (EEG) signals to evaluate human emotion. To attack different machine learning classifiers such as Adaptive Boosting (AdaBoost) and Random Forest dedicated to the classification of 4 different human emotions using EEG signals, this paper proposes two scenarios of label-flipping methods. The results of the studies show that the proposed data poison attacksm based on label-flipping are successful regardless of the model, but different models show different degrees of resistance to the assaults. In addition, numerous Explainable Artificial Intelligence (XAI) techniques are used to explain the data poison attacks on EEG signal-based human emotion evaluation systems.