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.

Large language models (LLMs) have attracted considerable attention as they are capable of showcasing impressive capabilities generating comparable high-quality responses to human inputs. LLMs, can not only compose textual scripts such as emails and essays but also executable programming code. Contrary, the automated reasoning capability of these LLMs in performing statistically-driven descriptive analysis, particularly on user-specific data and as personal assistants to users with limited background knowledge in an application domain who would like to carry out basic, as well as advanced statistical and domain-specific analysis is not yet fully explored. More importantly, the performance of these LLMs has not been compared and discussed in detail when domain-specific data analysis tasks are needed. Additionally, the use of LLMs in isolation is often at times insufficient for creating powerful applications and the real potential comes when LLMs are combined with other sources of computation such as LangChain. This study, consequently, explores whether LLMs can be used as generative AI-based personal assistants to users with minimal background knowledge in an application domain infer key data insights. To demonstrate the performance of the LLMs, the study reports a case study through which descriptive statistical analysis, as well as Natural Language Processing (NLP) based investigations, are performed on a number of phishing emails with the objective of comparing the accuracy of the results generated by LLMs to the ones produced by analysts. The experimental results show that LangChain and the Generative Pre-trained Transformer (GPT-4) excel in numerical reasoning tasks i.e., temporal statistical analysis, achieve competitive correlation with human judgments on feature engineering tasks while struggle to some extent on domain specific knowledge reasoning, where domain-specific knowledge is required.

There exist several data-driven approaches that enable us model time series data including traditional regression-based modeling approaches (i.e., ARIMA). Recently, deep learning techniques have been introduced and explored in the context of time series analysis and prediction. A major research question to ask is the performance of these many variations of deep learning techniques in predicting time series data. This paper compares two prominent deep learning modeling techniques. The Recurrent Neural Network (RNN)-based Long Short-Term Memory (LSTM) and the convolutional Neural Network (CNN)-based Temporal Convolutional Networks (TCN) are compared and their performance and training time are reported. According to our experimental results, both modeling techniques per-form comparably having TCN-based models outperform LSTM slightly. Moreover, the CNN-based TCN model builds a stable model faster than the RNN-based LSTM models.

It is important and informative to compare and contrast major economic crises in order to confront novel and unknown cases such as the COVID-19 pandemic. The 2006 Great Recession and then the 2019 pandemic have a lot to share in terms of unemployment rate, consumption expenditures, and interest rates set by Federal Reserve. In addition to quantitative historical data, it is also interesting to compare the contents of Federal Reserve statements for the period of these two crises and find out whether Federal Reserve cares about similar concerns or there are some other issues that demand separate and unique monetary policies. This paper conducts an analysis to explore the Federal Reserve concerns as expressed in their statements for the period of 2005 to 2020. The concern analysis is performed using natural language processing (NLP) algorithms and a trend analysis of concern is also presented. We observe that there are some similarities between the Federal Reserve statements issued during the Great Recession with those issued for the 2019 COVID-19 pandemic.

The detection of zero-day attacks and vulnerabilities is a challenging problem. It is of utmost importance for network administrators to identify them with high accuracy. The higher the accuracy is, the more robust the defense mechanism will be. In an ideal scenario (i.e., 100% accuracy) the system can detect zero-day malware without being concerned about mistakenly tagging benign files as malware or enabling disruptive malicious code running as none-malicious ones. This paper investigates different machine learning algorithms to find out how well they can detect zero-day malware. Through the examination of 34 machine/deep learning classifiers, we found that the random forest classifier offered the best accuracy. The paper poses several research questions regarding the performance of machine and deep learning algorithms when detecting zero-day malware with zero rates for false positive and false negative.

Machine and deep learning-based algorithms are the emerging approaches in addressing prediction problems in time series. These techniques have been shown to produce more accurate results than conventional regression-based modeling. It has been reported that artificial Recurrent Neural Networks (RNN) with memory, such as Long Short-Term Memory (LSTM), are superior compared to Autoregressive Integrated Moving Average (ARIMA) with a large margin. The LSTM-based models incorporate additional “gates” for the purpose of memorizing longer sequences of input data. The major question is that whether the gates incorporated in the LSTM architecture already offers a good prediction and whether additional training of data would be necessary to further improve the prediction. Bidirectional LSTMs (BiLSTMs) enable additional training by traversing the input data twice (i.e., 1) left-to-right, and 2) right-to-left). The research question of interest is then whether BiLSTM, with additional training capability, outperforms regular unidirectional LSTM. This paper reports a behavioral analysis and comparison of BiLSTM and LSTM models. The objective is to explore to what extend additional layers of training of data would be beneficial to tune the involved parameters. The results show that additional training of data and thus BiLSTM-based modeling offers better predictions than regular LSTM-based models. More specifically, it was observed that BiLSTM models provide better predictions compared to ARIMA and LSTM models. It was also observed that BiLSTM models reach the equilibrium much slower than LSTM-based models.

Forecasting time series data is an important subject in economics, business, and finance. Traditionally, there are several techniques to effectively forecast the next lag of time series data such as univariate Autoregressive (AR), univariate Moving Average (MA), Simple Exponential Smoothing (SES), and more notably Autoregressive Integrated Moving Average (ARIMA) with its many variations. In particular, ARIMA model has demonstrated its outperformance in precision and accuracy of predicting the next lags of time series. With the recent advancement in computational power of computers and more importantly development of more advanced machine learning algorithms and approaches such as deep learning, new algorithms are developed to analyze and forecast time series data. The research question investigated in this article is that whether and how the newly developed deep learning-based algorithms for forecasting time series data, such as "Long Short-Term Memory (LSTM)", are superior to the traditional algorithms. The empirical studies conducted and reported in this article show that deep learning-based algorithms such as LSTM outperform traditional-based algorithms such as ARIMA model. More specifically, the average reduction in error rates obtained by LSTM was between 84 - 87 percent when compared to ARIMA indicating the superiority of LSTM to ARIMA. Furthermore, it was noticed that the number of training times, known as "epoch" in deep learning, had no effect on the performance of the trained forecast model and it exhibited a truly random behavior.

Empirical studies show that coverage-based fault localizations are very effective in testing and debugging software applications. It is also a commonly held belief that no software testing techniques would perform best for all programs with various data structures and complexity. An important research question posed in this paper is whether the type and complexity of faults in a given program has any influence on the performance of these fault localization techniques. This paper investigates the performance of coverage-based fault localizations for different types of faults. We explore and compare the accuracy of these techniques for two large groups of faults often observed in object-oriented programs. First, we explore different types of traditional method-level faults grouped into six categories including those related to arithmetic, relational, conditional, logical, assignment, and shift. We then focus on class-level faults related to object-oriented features and group them into four categories including inheritance, overriding, Java-specific features, and common programming mistakes. The results show that coverage-based fault localizations are less effective for class-level faults associated with object-oriented features of programs. We therefore advocate the needs for designing more effective fault localizations for debugging object-oriented and class-level defects.

We introduce a technique to formally represent and specify race conditions in multithreaded applications. Answer set programming (ASP) is a logic-based knowledge representation paradigm to formally express belief acquired through reasoning in an application domain. The transparent and expressiveness representation of problems along with powerful non-monotonic reasoning power enable ASP to abstractly represent and solve some certain classes of NP hard problems in polynomial times. We use ASP to formally express race conditions and thus represent potential data races often occurred in multithreaded applications with shared memory models. We then use ASP to generate all possible test inputs and thread interleaving, i.e. scheduling, whose executions would result in deterministically exposing thread interleaving failures. We evaluated the proposed technique with some moderate sized Java programs, and our experimental results confirm that the proposed technique can practically expose common data races in multithreaded programs with low false positive rates. We conjecture that, in addition to generating threads scheduling whose execution order leads to the exposition of data races, ASP has several other applications in constraint-based software testing research and can be utilized to express and solve similar test case generation problems where constraints play a key role in determining the complexity of searches.