Neuro-Agentic Control: A Deep Learning-based LLM-Powered Agentic AI Framework for Controlling Security Controls

2026-07-09

AI Summary

This work addresses the limitations of traditional rule-based monitoring in industrial IoT systems, which struggle to counter sophisticated cyberattacks, and the unsuitability of large language models (LLMs) for closed-loop control due to their propensity for hallucination. To bridge this gap, the authors propose a neuro-agent control framework that integrates an LLM-based planner with TimesFM, a pretrained time-series foundation model. The framework simulates intervention effects within TimesFM’s latent numerical space and incorporates a “counterfactual physical injection” mechanism, leveraging the foundation model as a deterministic sentinel to filter out unsafe or physically infeasible actions. Evaluated on the SWaT dataset, the approach successfully prevents 33.3% of attacks from breaching critical thresholds—outperforming LSTM (26.7%) and TCN (13.3%) baselines—while executing zero physically invalid actions.

Abstract

Cyberattacks on operational technology are increasingly causing costly downtime and physical damage, exposing the limitations of traditional rule-based monitoring in industrial IoT environments. While Large Language Models (LLMs) have strong semantic reasoning abilities to assist in decision support, their hallucinatory nature presents unacceptable safety liabilities for closed-loop control. This paper introduces a neuro-agentic control framework, a novel architecture that couples an LLM-based planner (i.e., such as Gemini 2.5 Flash-Lite) with a pre-trained Time-Series Foundation Model (TimesFM), to achieve physics-grounded autonomous defense. The paper introduces a ``Counterfactual Physics Injection'' mechanism that simulates the impact of LLM-proposed interventions within the numerical latent space of the foundation model before actuation, while allowing the system to reject hallucinatory or unsafe actions. Evaluated on an industrial dataset (e.g., the Secure Water Treatment (SWaT)) in the context of stochastic attack scenarios, the framework exhibited better performance compared to LSTM and TCN baselines. The Neuro-Agentic Loop prevented five breaches (33.3%) below the threshold versus LSTM (26.7%) and TCN (13.3%), with zero physically invalid (hallucinated) actions executed. These results demonstrate the efficacy of using foundation models as deterministic ``Sentinels'' to safeguard agentic AI in critical infrastructure.

Problem

• Operational Technology Security
• Autonomous Defense
• LLM Hallucination
• Industrial IoT
• Cyber-Physical Systems

Innovation

• Neuro-Agentic Control
• Counterfactual Physics Injection
• Time-Series Foundation Model
• Hallucination Mitigation
• Autonomous Cyber Defense