Assessment of Deep Learning Models for Human Activity Recognition on Multi-Variate Time Series Data and Non-Targeted Adversarial Attack

Compares Keras-LSTM, RNN-LSTM, CNN, and ResNet classifiers on a feature-engineered human-activity-recognition dataset, with ResNet reaching 99.9% accuracy. Also evaluates model robustness under FGSM and Basic Iterative Method adversarial attacks, highlighting security risks in HAR deployment pipelines that had not previously been studied for this dataset and model family.

Recommended citation: Tasmin, M., Ruman, S.U., Ishtiak, T., Suhan, A.C., Hasif, R., Zulminan, S., Rahman, R.M. Assessment of Deep Learning Models for Human Activity Recognition on Multi-variate Time Series Data and Non-targeted Adversarial Attack. In: Advances in Intelligent Systems Research and Innovation, SSDC vol. 379. Springer (2022).
Download Paper