Subrat Swain

IIT Delhi, New Delhi, India

The University of Queensland, Brisbane, Australia

qiz218247@iitd.ac.in

I am a PhD researcher jointly supervised by Prof. Dan Kim at The University of Queensland and Prof. Vireshwar Kumar at IIT Delhi. My research lies at the intersection of adversarial machine learning and network security, with a focus on developing and defending against adversarial attacks on network intrusion detection systems, IoT networks, and other security-critical applications.

Prior to my PhD, I worked as a Product Engineer at Cognizant Technology Solutions, where I developed an AutoML-based ticket analysis solution. Before that, I was a Machine Learning Engineer at TaiyoAI Inc. (San Francisco, Remote), where I built Bayesian hyperparameter optimization systems and managed deployment pipelines for large-scale time series forecasting models.

I completed my Bachelor of Technology in Computer Science from VSS University of Technology, Odisha, with a CGPA of 9.21/10. I am a recipient of the Prime Minister’s Research Fellowship (PMRF) and ranked All India 800 (top 0.8%) in GATE CS-IT.

news

Jan 15, 2026	Invited speaker at the Academia-Industry Research Partnerships in Responsible AI Across Australia and India - UQ-IITD Industry Connect Workshop Series, 2026.
Oct 14, 2025	Attended ACM CCS 2025 at Taipei, Taiwan.
Jun 24, 2024	Paper PANDA: Practical Adversarial Attack Against Network Intrusion Detection accepted at DSN 2024 [CORE A].
Oct 01, 2023	Paper SPAT: Semantic-Preserving Adversarial Transformation for Perceptually Similar Adversarial Examples accepted at ECAI 2023 [CORE A].
Mar 01, 2023	Invited student speaker at the CSRC conference at The University of Queensland.

selected publications

ECAI
SPAT: Semantic-Preserving Adversarial Transformation for Perceptually Similar Adversarial Examples

Subrat Kumar Swain, Vireshwar Kumar, Dan D. Kim, and 1 more author

In European Conference on Artificial Intelligence (ECAI), 2023

[CORE A]

Abs Bib

We propose SPAT, a semantic-preserving adversarial transformation that generates perceptually similar adversarial examples while preserving the semantic content of the original input.
@inproceedings{swain2023spat, title = {{SPAT}: Semantic-Preserving Adversarial Transformation for Perceptually Similar Adversarial Examples}, author = {Swain, Subrat Kumar and Kumar, Vireshwar and Kim, Dan D. and Bai, Guangdong}, booktitle = {European Conference on Artificial Intelligence (ECAI)}, year = {2023}, note = {[CORE A]} }
DSN
PANDA: Practical Adversarial Attack Against Network Intrusion Detection

Subrat Kumar Swain, Vireshwar Kumar, Guangdong Bai, and 1 more author

In International Conference on Dependable Systems and Networks (DSN), 2024

[CORE A]

Abs Bib

We present PANDA, a practical adversarial attack framework targeting network intrusion detection systems, demonstrating the real-world vulnerability of ML-based security systems.
@inproceedings{swain2024panda, title = {{PANDA}: Practical Adversarial Attack Against Network Intrusion Detection}, author = {Swain, Subrat Kumar and Kumar, Vireshwar and Bai, Guangdong and Kim, Dan D.}, booktitle = {International Conference on Dependable Systems and Networks (DSN)}, year = {2024}, note = {[CORE A]} }