AgenticTwin: An Agentic LLM Framework Integrated with Digital Twin for Anomaly Detection

Digital twins are increasingly used to monitor and simulate the behavior of cyber-physical systems. Even with skilled operators, interpreting anomalies detected within digital twin pipelines is challenging, as the sheer complexity and volume of raw sensor data make thorough analysis difficult. Recent advances in large language models (LLMs) offer promising capabilities for reasoning and explanation, yet their integration into digital twin-driven anomaly analysis remains underexplored. In this work, we propose AgenticTwin, an agentic framework that integrates LLM-driven reasoning with a digital twin-based anomaly detection pipeline. The framework grounds LLM-generated explanations in outputs from a digital twin-driven anomaly classifier and enables human operators to ask relevant natural-language questions about the system. Beyond the framework itself, we introduce a benchmark-oriented evaluation pipeline constructed over synthetic anomalies injected into a real-world weather sensor dataset, enabling controlled generation of operator queries over anomaly events. We further evaluate the feasibility of deploying lightweight, open-source LLMs for practical cyber-physical environments. Experimental results demonstrate that structured agent collaboration and knowledge-grounded reasoning improve diagnosis quality, contextual retrieval, and mitigation quality across diverse possible anomaly scenarios.

Beware of Discarding Used SRAMs: Information is Stored Permanently

Data recovery has long been a focus of the electronics industry for decades by security experts, focusing on hard disk recovery, a type of non-volatile memory. Unfortunately, none of the existing research, neither from academia, industry, or government, have ever considered data recovery from volatile memories. The data is lost when it is powered off, by definition. To the best of our knowledge, we are the first to present an approach to recovering data from a static random access memory. It is conventional wisdom that SRAM loses its contents whenever it turns off, and it is not required to protect sensitive information, e.g., the firmware code, secret encryption keys, etc., when an SRAM-based computing system retires. Unfortunately, the recycling of integrated circuits poses a severe threat to the protection of intellectual properties. In this paper, we present a novel concept to retrieve SRAM data as aging leads to a power-up state with an imprint of the stored values. We show that our proposed approaches can partially recover the previously used SRAM content. The accuracy of the recovered data can be further increased by incorporating multiple SRAM chips compared to a single one. It is impossible to retrieve the prior content of some stable SRAM cells, where aging shifts these cells towards stability. As the locations of these cells vary from chip to chip due to uncontrollable process variation, the same cell has a higher chance of being unstable or stable against aging in any of the chips, which helps us recover the content. Finally, majority voting is used to combine a set of SRAM chips' data to recover the stored data. We present our experimental result using commercial off-the-shelf SRAMs with stored binary image data before performing accelerated aging. We demonstrate the successful partial retrieval on SRAMs that are aged with as little as 4 hours of accelerated aging with 85C.