Adversarial Threats from Machine Learning to Power Systems (AIThreatPS)
Announcements
- The task force was introduced at the IEEE Power and Energy Society General Meeting 2026 in Montreal during AIPSCC Main Meeting on Tuesday (7/21/2026), 3pm - 5pm in Room 511A. A copy of my slides can be found here.
- Join the task force: Please signup using this link or reach out to the task force co-chairs via email (at the bottom of the page)
History
This Taskforce was approved and established in July 2026 and sponsored by the IEEE PES AI for Power Systems Coordinating Committee (AIPSCC).
Task Force Purpose
The task force will develop methodologies, frameworks, and benchmark resources as well as guidelines and standards in long terms for ensuring the secure, resilient, and trustworthy deployment of artificial intelligence, including large language models and other AI technique, in power system operations. The task force aims to assess cyber risks associated with AI-enabled grid applications and to design adversarial-resilient solutions capable of detecting, withstanding, and recovering from cyber-attacks. We plan to develop representative test cases, datasets, and cyber-physical simulation environments that capture realistic power system operations, including transmission, distribution, and distributed energy resource integration, along with associated threat scenarios. These resources will support the evaluation of AI robustness, reliability, and security under diverse operating conditions. The outcomes of this task force will enable stakeholders to guide the safe integration of AI into critical energy infrastructure. The proposed work will aim to provide standardized evaluation approaches and support the development of guidelines and best practices for secure AI adoption in power systems.
Proposed Scope and Activities
- Identify and categorize cyber threats to AI-enabled power systems with operational use cases, including adversarial attacks, data/model poisoning, AI-enabled cyber intrusions, and vulnerabilities arising from existing and evolving grid architectures.
- Identify gaps in existing methodologies, standards, and tools for securing AI systems in power grid operations and define requirements for resilient and trustworthy AI deployment.
- Develop, validate, and disseminate adversarial-resilient AI frameworks and methodologies for detection, mitigation, and recovery from cyber-attacks in power systems.
- Develop benchmark test cases, datasets, and cyber-physical simulation environments for evaluating the robustness and security of AI-enabled grid applications.
- Coordinate with the PES AI coordinating council, Working Group and other Task Forces to align efforts on test systems, cybersecurity, and AI applications, and support the adoption of proposed methods in relevant initiatives.
- Organize meetings, panels, and webinars at IEEE PES General Meeting and other venues to engage industry and academic stakeholders and solicit feedback.
- Summarize findings in technical reports, white papers, and IEEE publications, and contribute to standards and recommended practices related to AI security in power system operation.
Task Force Co-Chairs
Dr. Subhash Lakshminarayana, University of Warwick, UK
Email: Firstname DOT Lastname AT warwick DOT ac DOT uk
Dr. Anurag Srivastava, West Virginia University, USA
Email: Firstname DOT Lastname AT mail DOT wvu DOT edu