Arun Kumar Elengovan Explores Agentic AI at ICSISCET 2025

The 7th International Conference on Sustainable and Innovative Solutions for Current Challenges in Engineering and Technology (ICSISCET 2025) concluded its two-day event on November 1 and 2 at the Madhav Institute of Technology and Science in Gwalior, India. Organised by the Soft Computing Research Society, this annual gathering attracted researchers, academic faculty, and technology professionals from various countries to discuss pressing issues in engineering and applied technology.

Arun Kumar Elengovan delivered a keynote address focused on agentic AI, highlighting the need for new strategies in cyber defense as artificial intelligence continues to evolve. His presentation, titled Beyond Prompts: How Agentic AI Will Redefine Cyber Defense, explored the transition of AI-driven systems into a phase where traditional methods may no longer suffice.

Innovative Approaches to Cyber Defense

ICSISCET 2025 served as a platform for international research collaboration, featuring both in-person sessions and an online programme via Google Meet. This hybrid format facilitated participation from attendees worldwide. The conference’s structured review process for submitted papers underscores its credibility within the research community.

In his keynote, Elengovan defined AI agents as systems capable of reasoning, action, and learning from interactions. He emphasized the transition from traditional automation to agent-based models, which allow for context-aware decision-making and dynamic responses to real-world conditions. Early in his presentation, he introduced large language model-based agents, which utilize reasoning engines to produce outcomes beyond predefined logic.

Elengovan argued that conventional detection-based strategies are inadequate as cyber threats evolve. Rather than relying on static rules, agentic systems can interpret signals, plan responses, and adjust behaviors based on real-time data. He highlighted the ReAct pattern, which enables agents to combine reasoning, action, and feedback for iterative decision-making.

Building the Future of Cyber Security

The keynote also delved into the foundational components of agentic security systems, such as reasoning engines, external tool invocation, and memory frameworks. Elengovan explained how short-term memory maintains context during tasks, while long-term memory enables agents to learn from previous interactions. He discussed the importance of orchestration in coordinating reasoning, action, and observation, which enhances system performance over time.

A significant segment of the presentation focused on the Model Context Protocol (MCP), a standardised method that allows agents to interact seamlessly with external tools. MCP simplifies the integration of multiple tools, enhancing reusability and enabling the addition of new capabilities across various agent systems. This approach minimizes operational friction and promotes scalability in cyber defense frameworks.

Elengovan also examined distributed agent frameworks such as DAPR Agents, LangChain, AutoGen, and CrewAI. These frameworks facilitate multi-agent collaboration and structured tool orchestration, allowing teams of agents to manage complex cyber defense tasks while maintaining quality control.

With a distinguished career in security engineering, Elengovan serves as the Director of Engineering Security at Okta, Inc.. His expertise spans applied cryptography, secrets management, and cyber defense strategies. He is recognised as a Gartner Information Security Ambassador and has reviewed over 180 research papers for esteemed journals and conferences.

The keynote resonated with attendees across multiple disciplines, offering a forward-looking perspective on the challenges and opportunities presented by autonomous AI systems. Elengovan’s insights grounded in practical engineering patterns connected advanced research to the operational realities of modern digital environments.

In closing, he encouraged the research and engineering communities to view agentic AI as a viable solution for securing future systems, rather than a speculative concept. As autonomous and distributed systems gain traction, the discussions at ICSISCET 2025 highlighted the growing necessity for adaptable, intelligent, and resilient defense mechanisms. Through this keynote, Arun Kumar Elengovan contributed significantly to the ongoing global dialogue on agentic AI and cyber defense.