This study, via some possible contributions, suggests an effective CTI and CTH framework to improve critical infrastructure security by creating a threat correlation engine, developing new CTI methods, and implementing a multi-view and multi-kernel deep learning system for CTH.

This paper provides new perspectives on explainable machine learning and deep learning, from the view point of an explicit system-centric approach. Unlike most existing commentary on the said topic, this paper seeks to explain the fundamentals of the black box problem in machine learning by focusing on explicit system modeling. A convenient yet interesting example is used to illustrate the ideas in a tutorial-like manner thereby making the article accessible to readers from diverse backgrounds.

Recurrence-based studies are employed for determining dynamical properties and for analyzing and processing time-series. Such studies rely on the so-called recurrence matrix, which is an orbit-specific binary representation of a proximity relation on the phase space. Customarily, this state proximity relation is deduced from the underlying metric structure via thresholding, and hence, a single user parameter is required for forming a recurrence matrix, unless embeddings are used. Since most of the prescriptions that can be found in the literature for identifying a suitable recurrence threshold are either heuristic or task-specific, a scaling-analysis framework, which identifies critical scales that emerge during the evolution based on the entropy of maximum sojourn time probability vectors, is introduced and validated here.

The increasing penetration of power-electronics interfaced resources brings new challenges regarding the small-signal stability of power systems. To address this issue, grid-forming controlled converters have emerged as an alternative to their conventional grid-following counterparts. This paper investigates the mechanisms behind converters driven stability and quantifies the stabilizing effect of grid-forming controls. The linearized state space model of different combinations of control strategies is analyzed in a multi-infeed system considering various operating points. Through a parametric sensitivity study and an examination of the participation factors of key eigenvalues of the linearized models, it is confirmed that grid-forming controls contribute to system stabilization. Moreover, this paper demonstrates that this stabilizing effect varies significantly depending on the specific grid-forming control implemented: whether a current control loop is used or not, notably impacts stability.

According to the World Health Organization (WHO), hundreds of million people have some type of visual disability. Vision impairment has a personal impact with lifelong consequences because more than 80% of our perception, cognition, learning, and daily activities are mediated through vision. Moreover, in the era of rapid advancements in artificial intelligence (AI), visually impaired and blind people face challenges at work and in education because of inaccessibility to AI technologies. In this regard, we present an assistive mobile application with an intuitive user interface (UI) for visually impaired and blind people to interact with ChatGPT via natural conversation. The app employs automatic speech recognition (ASR), text-to-speech (TTS), keyword spotting (KWS), voice activity detection (VAD), and a convenient UI to interact with ChatGPT effortlessly. We have made the source code, pre-trained models, and UI publicly available at https://github.com/IS2AI/talk-llm to stimulate the development of assistive mobile applications.

In recent decades, the structure of the mod-2 cohomology of the Steenrod ring $\mathscr A$ has become a major subject of study in the field of Algebraic Topology. One of the earliest attempts to study this cohomology through the use of modular representations of the general linear groups was the groundbreaking work [Math. Z. \textbf{202} (1989), 493-523] by W.M. Singer. In that work, Singer introduced a homomorphism, commonly referred to as the “algebraic transfer,” which maps from the coinvariants of a certain representation of the general linear group to the mod-2 cohomology group of the ring $\mathscr A.$ Singer’s conjecture, in particular, which states that the algebraic transfer is a monomorphism for all homological degrees, remains a highly significant and unresolved problem in Algebraic Topology. In this research, we take a major stride toward resolving the Singer conjecture by establishing its truth for the homological degree four.

Machine learning has the potential to improve the accuracy of period tracking applications by analyzing patterns in menstrual cycle data. The ability to predict the timing of menstrual cycles is important for women's health, and can be used to provide personalized reminders and notifications, as well as to better understand and manage menstrual cycles.

The following is a conversation with ChatGPT, a large language machine learning model, on the topic of quantum. In the discussion section, the responses of ChatGPT are analyzed. Although comprehensive, impressive, and full of surprises, ChatGPT is not yet free of errors, typos, or obscure suggestions. A particular recommendation is to include awareness to ChatGPT about the source of its knowledge.

The connected and autonomous vehicles   (CAV)   applications and services-based traffic make an extra burden on the already congested cellular networks. Offloading is envisioned as a promising solution to tackle cellular networks’  traffic explosion problem. Notably, vehicular traffic offloading leveraging different vehicular communication network (VCN) modes is one of the potential techniques to address the data traffic problem in cellular networks. This paper surveys the state-of-the-art literature for vehicular data offloading under a communication perspective,  i.e.,  vehicle to vehicle  (V2V),  vehicle to roadside infrastructure  (V2I),  and vehicle to everything  (V2X). First, we pinpoint the significant classification of vehicular data/traffic offloading techniques, considering whether data is to download or upload. Next, for better intuition of each data offloading’s category,  we sub-classify the existing schemes based on their objectives. Then, the existing literature on vehicular data/traffic is elaborated, compared, and analyzed based on approaches, objectives, merits, demerits, etc. Finally, we highlight the open research challenges in this field and predict future research trends.

Electromagnetic forward and inverse solvers are considered for the analysis and design of electromagnetic metasurfaces. This paper reviews (i) a forward solver that takes the surface susceptibilities of metasurfaces and calculates their scattered fields and (ii) an inverse solver that takes user-defined desired power patterns and outputs the required metasurface properties. An omega-bianisotropic transmitting metasurface is then fabricated and measured to evaluate the performance of these forward and inverse solvers.

Resubmitted for review at IEEE TMI on February 7th, 2023.

This study describes several critical security conditions and requirements, as well as security threats, related to smart farming and precision agriculture environments.

Deep Learning (DL) has become a popular method in various industries to support human decision-making in areas such as gaming, news, finance, biology, medicine, and engineering. Although statistical learning has been used in clinical decision support for some time, deep learning has not been widely implemented in the clinical field, apart from image-related tasks. Clinical natural language processing (NLP) in decision support systems continues to rely mostly on rule-based or statistical solutions, unlike other domains that have adopted deep learning. The development of clinical NLP applications in languages other than English, such as Japanese, is still lagging behind. This paper offers a review on the challenges that deep learning faces in clinical NLP and explores potentials and opportunities in the field. The purpose is to provide a broad understanding of the clinical NLP landscape for deep learning researchers and inspire them to get involved in this field.

Deep learning has impacted various aspects of artificial intelligence ranging from computer vision to reinforcement learning. However, it has also inspired malicious actors to spread misinformation without detection. Fake videos can now be generated with precision high enough to go undetected by traditional video forensics. These videos, popularly referred to as DeepFakes, are being used to spread misinformation. Deep learning based techniques have evolved over the years to detect such DeepFakes. This article analyzes the DeepFake detection algorithms and implements five such algorithms on a standard dataset. The advantages and challenges of the implemented algorithms are explored in detail.

In recent years, the push for accurate and reliable time synchronization has become increasingly important in crit?ical infrastructure, particularly in telecommunication networks. The enhanced performance of 5G New Radio and next-generation technologies rely on phase synchronization of Radio Access Network (RAN) nodes, which require sub-microsecond relative timing errors. Atomic clocks, integrated with Global Navigation Satellite Systems (GNSS) timing receivers, have been deployed in timing networks as Grand Master Clocks (GMCs). However, this solution does not scale well with the growing number of interme?diate nodes in current RANs. A more affordable and distributed solution is needed for scalability and time synchronization. GNSS timing receivers are a cost-effective solution providing stable reference clock signals, but a proliferation of GNSS antennas can expose the network to malicious radio-frequency attacks. This research proposes a solution for stable and resilient GNSS?based network synchronization, using the White Rabbit Precise Time Protocol and a timing source backup logic in case of timing-disruptive attacks. The solution was tested against popular jamming, meaconing, and spoofing attacks and was able to maintain 2 ns relative synchronization accuracy between its nodes under any of the tested attacks, without the support of an atomic clock.

The prediction of energy consumption of users is an important requirement for utilities. Machine learning using neural networks has been applied for predicting household energy consumption due to effectiveness in capturing correlations and patterns. However, the result is predicted by learning over the behavior of the training data as a whole and ignoring the differences between sub-group behavior (i.e., groups of users with similar characteristics). The prediction of energy consumption of users is an important requirement for utilities. Exploiting the behavioral differences between different subgroups in the population can improve the system’s capability to perform parallel processing of similar tasks. This paper proposes an innovative model to simultaneously predict gas consumption of multiple households or population subgroups. The proposed method is based on the use of data transformations and focuses on differences in individual behavior. The evaluation of the model is discussed from three aspects: prediction accuracy, risk of privacy, as well as system robustness. The results shows that the proposed system can achieve multiple prediction simultaneously with reasonable accuracy and be secure against noise and attacks on private information.

Photovoltaics (PV) in onboard vehicle applications adds weight to an electric vehicle (EV), increasing the overall energy consumption. Although the added PV system weight (1.5– 40 kg) is small compared to the vehicle weight (1500–2200 kg), the power generated by PV (55–700 W) is also very small com- pared to the power needed (up to 80–285 kW) to propel an EV, making the effect of additional PV system weight on energy con- sumption a non-trivial topic to analyze. We present a method to study the impact of vehicle onboard PV weight on the energy balance of EVs for different Vehicle-added PV (VAPV) and vehi- cle-integrated PV (VIPV) configurations with eight different PV technologies, using data from vehicle onboard measurement campaigns and simulations. Simulations are carried out for the driving phase of two electric cars (medium and large passenger cars). Our method calculates the energy consumption attributa- ble to the added PV system weight (0.05–1.4 Wh/km) and PV energy yield (0.12–3.12 Wh/km) for a selection of trips. The re- sults of these simulations are expressed through a newly intro- duced parameter called “onboard PV yield factor”, where posi- tive values indicate a net energy gain and negative values indicate a net energy loss of the onboard PV system. Our results show that the onboard PV yield factor for a VAPV configuration can range between -69.1 and 86.9 %, and for a VIPV configuration, between 77.2 and 89.7 %.

A review of Secure Constrained Node Networks and a performamce evaluation using traffic analysis in a Cooja based RPL: IPv6 Routing Protocol for Low-Power and Lossy Network.