Recent advancements in Large Language Models (LLMs) have revolutionized Natural Language Processing (NLP) and Natural Language Understanding (NLU), showcasing their ability to produce coherent and contextually relevant responses. However, their widespread use raises serious concerns regarding the potential for generating and spreading false or misleading information. This research examines the effectiveness of customized Retrieval Augmented Generation (RAG) models, alongside fine-tuned versions of GPT-Neo and RoBERTa.The proposed framework leverages multiple generative language models, including GPT-Neo, RoBERTa, and a custom RAG model, to produce diverse fake news content grounded in retrieved contextual information. It employs a combination of large language models and a specialized fake news detection pipeline, which integrates embedding based retrieval with sentence transformers and Facebook AI Similarity Search (FAISS), while also enabling generation through GPT-Neo, RoBERTa, and custom RAG structures. Additionally, we employ a passive aggressive classifier trained on "Fake" and "Real" dataset from a public GitHub repository to assess the likelihood of generated responses being classified as "Fake" or "Real." This pipeline evaluates the authenticity of news articles and incorporates believability scores to enhance interpretability.Results indicate that while all models perform comparably, the custom RAG model consistently excels in providing contextually grounded and highly relevant fake information,in these fake news scenarios. This study highlights the robustness of retrieval augmented frameworks in adversarial tasks, offering superior alignment with factual references. It contributes to the AI driven misinformation detection landscape, providing valuable insights into model selection and training methodologies to combat social engineering and the spread of fake news.

Information plays a key role in decision making, influencing others and risk assessment. However, malicious actors often falsify facts to mislead viewers. With the advent of Generative AI models, creating and modifying information has become much easier and more accessible, significantly reducing the time and effort required for attackers. To address the challenges of misinformation and disinformation, this paper presents the "Information Reliability Framework (IRF)," a detection framework built with DarkBERT (a dark web-focused language model) [1] and LLaMA-3.1 (a large language model), which incorporates real-time web data for validation. DarkBERT was fine-tuned using the FEVER dataset (Fact Extraction and VERification) [2] to enhance its efficiency in misinformation, disinformation, and fact-checking tasks. This fine-tuning enables DarkBERT to assign a reliability score to a given statement by analyzing malicious keywords and contexts, which is then combined with LLaMA-3.1 to generate a detailed explanation for the user, leveraging its knowledge and logical reasoning. To minimize reliance on outdated information and reduce hallucinations in LLaMA-3.1, the framework integrates real-time web data using Retrieval-Augmented Generation (RAG). The framework was evaluated alongside standalone LLaMA-3.1 with RAG using two synthetic datasets: generic statements and DarkBERT-specific statements, generated by ChatGPT-4o with its latest "search the web" feature. The evaluation showed that the IRF provided promising results across metrics such as accuracy, precision, recall, and F1 score, making it effective for validating text from various sources, including social media, emails, articles, messages, websites, and blogs.

Ethereum Smart Contract (SC) is a sophisticated technology that enhances the scope of BlockChain automation. However, SCs’ vulnerable programs have marred BlockChain’s nascent technology’s brighter aspects. Two critical hacks, the DAO attack, caused by reentrancy vulnerability, and the Parity attack, caused by unprotected selfdestruct and frozen Ether vulnerabilities, are famous for their historical cryptocurrency frauds. DAO attack robbed a sixty million dollar amount of cryptocurrency from the victim’s account. But the Parity attack created a new trend in software vulnerabilities by freezing a thirty million dollar amount of Ether. In fact, the Parity attack wiped out the library SC. Subsequently, all the SCs, depending upon the library functions (e.g., the Parity SC’s Ether transfer), became paralyzed, which locked the investors’ funds. This paper enhances our dynamic analysis tool, TechyTech, to detect the Locked Ether (i.e., Frozen Ether) vulnerability. Furthermore, TechyTech adopts a transfer-based approach and uses case studies to compare TechyTech’s performance with Remix.

Phishing websites many a times look-alike to benign websites with the objective being to lure unsuspecting users to visit them. The visits at times may be driven through links in phishing emails, links from web pages as well as web search results. Although the precise motivations behind phishing websites may differ the common denominator lies in the fact that unsuspecting users are mostly required to take some action e.g., clicking on a desired Uniform Resource Locator (URL). To accurately identify phishing websites, the cybersecurity community has relied on a variety of approaches including blacklisting, heuristic techniques as well as content-based approaches among others. The identification techniques are every so often enhanced using an array of methods i.e., honeypots, features recognitions, manual reporting, web-crawlers among others. Nevertheless, a number of phishing websites still escape detection either because they are not blacklisted, are too recent or were incorrectly evaluated. It is therefore imperative to enhance solutions that could mitigate phishing websites threats. In this study, the effectiveness of the Bidirectional Encoder Representations from Transformers (BERT) is investigated as a possible tool for detecting phishing URLs. The experimental results detail that the BERT transformer model achieves acceptable prediction results without requiring advanced URLs feature selection techniques or the involvement of a domain specialist.

The advent of the Internet has enabled everyone with access to it to provide their views online. This freedom of expression has also resulted in an increasing amount of unstructured text data daily, which can be leveraged to build models that can help make better business decisions. Customer reviews have become an integral part of the decision-making process as there is a tremendous increase in online products and services. Reviews provided by users online have a major problem regarding reliability and authenticity. It is an arduous task to make business decisions based on unstructured reviews whose trustworthiness is not established. Hence, this paper focuses on classifying the reviews of certain restaurants available on the Internet using different machine/deep learning techniques and summarises the findings. The results show that deep learning methods are more efficient in identifying fake reviews. More specifically, combining BERT and a 4-layered Feed Forward network gave 96% accuracy in detecting fake reviews.

With the recent advent of Large Language Models (LLMs), such as ChatGPT from OpenAI, BARD from Google, Llama2 from Meta, and Claude from Anthropic AI, gain widespread use, ensuring their security and robustness is critical. The widespread use of these language models heavily relies on their reliability and proper usage of this fascinating technology. It is crucial to thoroughly test these models to not only ensure its quality but also possible misuses of such models by potential adversaries for illegal activities such as hacking. This paper presents a novel study focusing on exploitation of such large language models against deceptive interactions. More specifically, the paper leverages widespread and borrows well-known techniques in deception theory to investigate whether these models are susceptible to deceitful interactions. This research aims not only to highlight these risks but also to pave the way for robust countermeasures that enhance the security and integrity of language models in the face of sophisticated social engineering tactics. Through systematic experiments and analysis, we assess their performance in these critical security domains. Our results demonstrate a significant finding in that these large language models are susceptible to deception and social engineering attacks.

This paper presents a study of irradiance prediction using a transformer-based machine learning model for the photovoltaic (PV) renewable energy system. We explore the forecast of irradiance using ten years of data from the Texas Mesonet Data Archive at the Reese Center in Lubbock, Texas. After training our transformer model with 90% of the data, we show that it can fit the irradiance trend successfully, while the testing phase with the remaining 10% of the dataset indicates that the transformer can predict irradiance trends that align with the observed values. Additionally, we borrow the concept of rolling LSTMs to generate a rolling transformer in order to extrapolate values of irradiance even when the observed values are not available. Our extrapolation results show that the transformer can extrapolate irradiance values accurately in the short-term, but it is less precise in the long-term. To remedy this, we aim to explore more thoroughly the hyperparameter configuration of our model in order to move towards our goal of including machine learning methods in the control of PV systems.

This paper extends the applications of ContextMiner, a framework that we proposed recently aiming to extract interpretable contextual features to conceptualize knowledge in security texts. We show that contextual features are suitable for building a document-level representation that can be used further in a downstream machine learning and natural language processing task: document clustering. Our results show that the intrinsic readability of contextual features is usable for analyzing the document clusters obtained in our experiments. Such analysis is performed through a density-based feature selection, cluster visualization, and statistical methods in order to unveil which contextual features better characterize each document cluster. Our findings suggest that statistical techniques alongside feature analysis can be utilized to discover meaningful commonalities among documents in a particular cluster without the need of querying each document manually.

It is important to learn about attackers and their attacking strategies so that better and more effective defense systems can be built. During the reconnaissance stage, attackers intend to probe potential targets through various techniques including social engineering attacks. Phishing through email is a well-known, cheap, easy, and surprisingly effective technique for obtaining the needed information. This type of attack targets individuals and thus utilizes weaknesses that might exist in each person. Given the uniqueness of each individual’s personality, attackers make sure the right persuasion principle technique is employed for each targeted individual. This paper describes efforts to build machine-learning transformers, the emerging technique in language modeling, with the goal of building classifiers that take into account different types of persuasion principles. More specifically, the paper describes efforts to build machine-learning transformers based on BERT, RoBERTa, and DistilBERT and captures their classification results. The results show that these transformers are accurate enough to build a classification of phishing emails with respect to persuasion techniques. Furthermore, we report that the RoBERTa model is able to train faster than BERT and DistilBERT models.

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.

The Internet of Things (IoT) has already changed our daily lives by integrating smart devices together towards delivering high quality services to its clients. These devices when integrated together form a network through which massive amount of data can be produced, transferred, and shared. A critical concern is the security and integrity of such a complex platform to ensure the sustainability and reliability of these IoT-based systems. Blockchain is an emerging technology that has demonstrated its unique features and capabilities for different problems and application domains including IoT-based systems. This survey paper reviews the adaptation of Blockchain in the context of IoT to represent how this technology is capable of addressing the integration and security problems of devices connected to IoT systems. The innovation of this survey is that we present a survey based upon the integration approaches and security issues of IoT data and discuss the role of Blockchain in connection with these issues.

Sonification is the science of communication of data and events to users through sounds. Auditory icons, earcons, and speech are the common auditory display schemes utilized in sonification, or more specifically in the use of audio to convey information. Once the captured data are perceived, their meanings, and more importantly, intentions can be interpreted more easily and thus can be employed as a complement to visualization techniques. Through auditory perception it is possible to convey information related to temporal, spatial, or some other context-oriented information. An important research question is whether the emotions perceived from these auditory icons or earcons are predictable in order to build an automated sonification platform. This paper conducts an experiment through which several mainstream and conventional machine learning algorithms are developed to study the prediction of emotions perceived from sounds. To do so, the key features of sounds are captured and then are modeled using machine learning algorithms using feature reduction techniques. We observe that it is possible to predict perceived emotions with high accuracy. In particular, the regression based on Random Forest demonstrated its superiority compared to other machine learning algorithms.

The growth of Cyber-physical Systems (CPS) has been increased in recent years. This has led to the coupling of highly complex cyber-physical components. With the integration of such complex components, new security challenges have emerged. Studies involving security issues in CPS have been quite difficult to be generalized due to the presence of heterogeneity and the diversity of the CPS components. These systems are subject to various vulnerabilities, threats and attacks, as a consequence of complex versions of CPS being introduced over time. This paper deals with vulnerabilities caused due to improper configurations in the software component of cyber-physical systems. Evolutionary algorithms such as Genetic Algorithms (GA) and Particle Swarm Optimization (PSO) can be employed to adequately test the underlying software for certain categories of vulnerabilities. This paper provides a detailed sensitivity analysis of these evolutionary algorithms in order to find out whether changing parameters involved in tuning these algorithms affect the overall performance. This analysis is based on the estimate of the number of generation of secure vulnerability pattern vectors under the variation of different parameters. The results indicate that while there is no evidence of influential parameters in Genetic Algorithms (i.e., mutation rate and population size), changes in the parameters involved in Particle Swarm Optimization algorithms (i.e., velocity rate and fitness range) have some positive impacts on the number of secure configurations generated.

Cyber-physical systems posit a complex number of security challenges due to interconnection of heterogeneous devices having limited processing, communication, and power capabilities. Additionally, the conglomeration of both physical and cyber-space further makes it difficult to devise a single security plan spanning both these spaces. Cyber-security researchers are often overloaded with a variety of cyber-alerts on a daily basis many of which turn out to be false positives. In this paper, we use machine learning and natural language processing techniques to predict the consequences of cyberattacks. The idea is to enable security researchers to have tools at their disposal that makes it easier to communicate the attack consequences with various stakeholders who may have little to no cybersecurity expertise. Additionally, with the proposed approach researchers' cognitive load can be reduced by automatically predicting the consequences of attacks in case new attacks are discovered. We compare the performance through various machine learning models employing word vectors obtained using both tf-idf and Doc2Vec models. In our experiments, an accuracy of 60% was obtained using tf-idf features and 57% using Doc2Vec method for models based on LinearSVC model.

The problem of detecting phishing emails through machine learning techniques has been discussed extensively in the literature. Conventional and state-of-the-art machine learning algorithms have demonstrated the possibility of building classifiers with high accuracy. The existing research studies treat phishing and genuine emails through general indicators and thus it is not exactly clear what phishing features are contributing to variations of the classifiers. In this paper, we crafted a set of phishing and legitimate emails with similar indicators in order to investigate whether these cues are captured or disregarded by email embeddings, i.e., vectorizations. We then fed machine learning classifiers with the carefully crafted emails to find out about the performance of email embeddings developed. Our results show that using these indicators, email embeddings techniques is effective for classifying emails as phishing or legitimate.

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.

Water is a critical natural resource that has significant impacts on human living and society. Growing population and energy consumption exacerbate the scarcity of water and our ability to manage this resource. This demonstration paper presents WaterScope, a smart and connected platform for water resource management, which integrates multiple data sources such as water level data, social media data, and water related articles. Furthermore, the tool enables forecasting underground water levels, identifying water concerns, sharing knowledge and expertise among stakeholders, and thus bringing new insights to our understanding and insights of the water supplies and resource management. The prototype engages water stakeholders who face problems of similar nature but deal with the problem in an ad-hoc and isolated manner. The interactive WaterScope platform targets creating an interconnected virtual community that aims to improve water supply resilience.

Billions of devices in the Internet of Things (IoT) are inter-connected over the internet and communicate with each other or end users. IoT devices communicate through messaging bots. These bots are important in IoT systems to automate and better manage the work flows. IoT devices are usually spread across many applications and are able to capture or generate substantial influx of big data. The integration of IoT with cloud computing to handle and manage big data, requires considerable security measures in order to prevent cyber attackers from adversarial use of such large amount of data. An attacker can simply utilize the messaging bots to perform malicious activities on a number of devices and thus bots pose serious cybersecurity hazards for IoT devices. Hence, it is important to detect the presence of malicious bots in the network. In this paper we propose an evidence theory-based approach for malicious bot detection. Evidence Theory, a.k.a. Dempster Shafer Theory (DST) is a probabilistic reasoning tool and has the unique ability to handle uncertainty, i.e. in the absence of evidence. It can be applied efficiently to identify a bot, especially when the bots have dynamic or polymorphic behavior. The key characteristic of DST is that the detection system may not need any prior information about the malicious signatures and profiles. In this work, we propose to analyze the network flow characteristics to extract key evidence for bot traces. We then quantify these pieces of evidence using apriori algorithm and apply DST to detect the presence of the bots.

With the continued improvement and innovation, technology has become an integral part of our daily lives. The rapid adoption of technology and its affordability has given rise to the Internet-of-Things (IoT). IoT is an interconnected network of devices that are able to communicate and share information seamlessly. IoT encompasses a gamut of heterogeneous devices ranging from a small sensor to large industrial machines. One such domain of IoT that has seen a significant growth in the recent few years is that of the wearable devices. While the privacy issues for medical devices has been well-researched and documented in the literature, the threats to privacy arising from the use of consumer wearable devices have received very little attention from the research community. This paper presents a survey of the literature to understand the various privacy challenges, mitigation strategies, and future research directions as a result of the widespread adoption of wearable devices.

The emerging Internet of Things (IoT) has increased the complexity and difficulty of network administration. Fortunately, Software-Defined Networking (SDN) provides an easy and centralized approach to administer a large number of IoT devices and can greatly reduce the workload of network administrators. SDN-based implementation of networks, however, has also introduced new security concerns, such as increasing number of DDoS attacks. This paper introduces an easy and lightweight defense strategy against DDoS attacks on IoT devices in a SDN environment using Markov Decision Process (MDP) in which optimal policies regarding handling network flows are determined with the intention of preventing DDoS attacks.