DARPA Program · 2023–Present

DARPA CASTLE

Cyber Agents for Security Testing and Learning Environments — Training autonomous AI agents to defend computer networks against advanced persistent threats. A $6.89M, four-year DARPA grant led by Vanderbilt University.

Program Overview

CASTLE develops a toolkit that instantiates realistic network environments and trains AI agents to perform cybersecurity assessments autonomously.

$6.89M
Grant Value
4 Year
Duration
3 Nations
US-UK-Canada Trilateral

I am a named researcher and co-author on the RAMPART project under CASTLE. My work focuses on developing reinforcement learning-based approaches for autonomous network defense and building realistic OT cybersecurity testbeds for training and evaluation.

Research Works

RAMPART: Reinforcement Against Malicious Penetration by Adversaries in Realistic Topologies

RAMPART: Reinforcement Against Malicious Penetration by Adversaries in Realistic Topologies

H. Neema, D. Balasubramanian, **H. Vardhan**, S. Neema
CPSIoTWeek 2024
RLCyber Defense
Autonomous Cybersecurity Testbed for Operational Technology Networks

Autonomous Cybersecurity Testbed for Operational Technology Networks

A. Raj, S. Das, **H. Vardhan**, H. Neema, A. Chhokra, D. Balasubramanian
DESTION@CPSIoTWeek 2025
OT SecurityTestbed
Cyber Security of OT Networks: A Tutorial and Overview

Cyber Security of OT Networks: A Tutorial and Overview

S. Kumar, **H. Vardhan**
arXiv 2025
OT/ICSSurvey

Media Coverage

Vanderbilt Engineering
December 12, 2023
DARPA Official
October 2022
DARPA — Trilateral Announcement
CASTLE featured in DARPA's first US-UK-Canada trilateral research partnership.
September 2024
Federal News Network
Discusses CASTLE's 'cyber gym' concept.
February 2024
Military Embedded Systems
December 2022
Homeland Security Today
October 2022
Executive Gov
September 2024
DOT&E (Department of Defense)
March 2023
CSIAC (DoD)
November 2022
MeriTalk
2022