The introduction of smart grids allows utility providers to collect detailed data about consumers, which can be utilized to enhance grid efficiency and reliability. However, this data collection also raises privacy concerns. To protect user privacy, some studies suggest using battery-based load hiding. Nevertheless, the impact of widespread adoption of this approach on utility providers remain unclear. Our paper seeks to evaluate the effects of battery-based load hiding on two critical operations: user profiling and anomaly detection. Our findings reveal that the inclusion of battery users in datasets can diminish the quality of conclusions drawn from these data. This can result in a decrease in the Area Under the Curve (AUC) by more than 10% when attempting to profile users within single-occupant and multiple-occupant households. Furthermore, our experiments demonstrate that battery-based load hiding not only conceals information about users employing the batteries but can also lead to an increased rate of false positives for other non-battery users (from 0.15 to 0.37) within the system. To mitigate these adverse effects, our study assessed various mitigation strategies. In the context of user profiling, our experiment demonstrated that identifying and removing battery users from the analytical dataset using unsupervised detection methods can effectively lessen the impact of battery users. For anomaly detection, our experiment revealed that creating separate classification models for battery and non-battery users can significantly reduce the adverse influence of battery users on the detection performance.

AbstractThe development of smart grids, traditional power grids, and the integration of internet of things devices have resulted in a wealth of data crucial to advancing energy management and efficiency. Nevertheless, public datasets remain limited due to grid operators' and companies' reluctance to disclose proprietary information. The authors present a comprehensive analysis of more than 50 publicly available datasets, organised into three main categories: micro‐ and macro‐consumption data, detailed in‐home consumption data (often referred to as non‐intrusive load monitoring datasets or building data) and grid data. Furthermore, the study underscores future research priorities, such as advancing synthetic data generation, improving data quality and standardisation, and enhancing big data management in smart grids. The aim of the authors is to enable researchers in the smart and power grid a comprehensive reference point to pick suitable and relevant public datasets to evaluate their proposed methods. The provided analysis highlights the importance of following a systematic and standardised approach in evaluating future methods and directs readers to future potential venues of research in the area of smart grid analytics.

This paper introduces a novel residual-based model to identify households with Battery Electric Vehicles (EVs) under high Air Conditioning (AC) load. The considerable energy demands of AC units can obscure charging events for EVs. In this work we propose a residual based model which leverages the distinctive characteristics of EV charging patterns, marked by unpredictable spikes in energy consumption, and the more predictable nature of AC load. Our proposed approach involves training a lightweight forecasting model to predict overall house-hold consumption and utilizes the residuals of this model for iden-tifying household with EVs. The residual-based model, ResEV-AR, demonstrated a substantial advantage in F1 score (5.8% and 7.32%) compared to state-of-the-art models such as EVS and KBF, respectively. Additionally, a simpler residual model, ResEV-SRM, exhibited a 3.5% F1 score advantage over EVS, coupled with an impressive 11-fold reduction in computation time.

Exploiting beamforming techniques to direct radio waves into narrow beams targeted at users are essential in 5G and 6G to ensure high data rate requirements with low latency. However, beamforming in a dynamic environment is challenging due to the high sensitivity to shift in the receiver location. Therefore, an adaptive beamforming approach is needed. In this paper, we utilize DeepSense 6G and 5G MIMO datasets for optimal beam prediction using deep learning approaches. We introduce a simple approach to transform radar input into vector embeddings in the DeepSense 6G dataset. Moreover, we address the severe minority classes problem in the 5G MIMO dataset. Finally, we conduct a thorough analysis of the state-of-the-art models. The proposed model shows significant improvements in the Top-1, Top-3, Top-5, weighted average and distance based accuracies.

Children during their early years might exhibit different forms of aggression against others. Aggression can be physical or relational. Physical aggression can take on different forms such as hitting, pushing, and kicking. The integration of technology, such as social robots, can be used to address aggression among children during childhood. In this study, we present a framework consisting of using sensory modules to detect undesirable physical interactions and social robots to provide a feedback once a behavior is detected. The framework has been demonstrated using a commercially-available social robot (i.e., Professor Einstein) with Raspberry Pi as the sensory module. Experiments with the social robot showed a promising performance of this integration. The outcomes can be used to teach children how to interact with others in an acceptable manner. The proposed framework can be integrated by social roboticists into their designs to create more dynamic interactions targeting aggressive and unwanted interactions.

Smart meters are continuously being deployed in several countries as a step in the direction of modernizing the power grid. Smart meters allow for automatic electricity consumption reporting to energy providers to facilitate billing and demand-based power generation. However, research has shown that such high resolution reporting to suppliers can potentially be used to invade consumers’ privacy; by identifying and predicting their behavior based on their consumption readings. This work presents a new protocol to preserve users’ privacy while maintaining the benefits of smart grids. The proposed method utilizes different techniques like randomization, masking, and differential privacy to build the scheme. The proposed method is shown to be more efficient compared to previous work in terms of performance and communication overhead. The implementation, simulation, and analysis are performed on datasets of real smart meters readings of households and electric vehicle chargers.

The Smart Grid Advanced Metering Infrastructure (AMI) has revolutionized the smart grid network, generating vast amounts of data that can be utilized for diverse objectives, one of which is Household Characteristics Classification (HCC). This can help the utility provider profile their customers and tailor their services to meet customer needs. To accomplish this task, we evaluated multiple Machine learning HCC models, with a focus on CNN-based models due to their wide popularity in the field of smart grid signal classification. We evaluated 1D and 2D variants of four different CNN architectures. Our experimental analysis revealed that ResNet-based models achieved the best performance on the task of HCC. Also, we found that 2D models tends to perform better than 1D variants.

Several fake claims are commonly repeated over time, especially on social media. To identify such previous claims, the verified claim retrieval task was studied, where, for a given input claim, the goal is to find previously-verified claims that are relevant to it. However, this view assumes that each claim was already verified, which may not be true for all claims in the real -world scenario. In this work, we introduce the Verified Claim Checking problem over Twitter, in which the relevant verified claims are retrieved only if the input claim was indeed previously -verified, thus saving computation time. We address the problem by proposing SpotVC, an end-to-end approach consisting of two stages, namely a filter and a reranker. The proposed filter achieved an average F1 of 0.81 while significantly reducing computation time. Moreover, the proposed reranker outperformed the state-of-the-art models on two public datasets and provided on-par performance on a third one. Overall, our proposed system exhibits an effective operational balance in the trade-off between efficiency and effectiveness for the real-world scenario.

Children with autism face challenges in various skills (e.g., communication and social) and they exhibit challenging behaviours. These challenging behaviours represent a challenge to their families, therapists, and caregivers, especially during therapy sessions. In this study, we have investigated several machine learning techniques and data modalities acquired using wearable sensors from children with autism during their interactions with social robots and toys in their potential to detect challenging behaviours. Each child wore a wearable device that collected data. Video annotations of the sessions were used to identify the occurrence of challenging behaviours. Extracted time features (i.e., mean, standard deviation, min, and max) in conjunction with four machine learning techniques were considered to detect challenging behaviors. The heart rate variability (HRV) changes have also been investigated in this study. The XGBoost algorithm has achieved the best performance (i.e., an accuracy of 99%). Additionally, physiological features outperformed the kinetic ones, with the heart rate being the main contributing feature in the prediction performance. One HRV parameter (i.e., RMSSD) was found to correlate with the occurrence of challenging behaviours. This work highlights the importance of developing the tools and methods to detect challenging behaviors among children with autism during aided sessions with social robots.

This paper investigates the potential privacy risks associated with forecasting models, with specific emphasis on their application in the context of smart grids. While machine learning and deep learning algorithms offer valuable utility, concerns arise regarding their exposure of sensitive information. Previous studies have focused on classification models, overlooking risks associated with forecasting models. Deep learning based forecasting models, such as Long Short Term Memory (LSTM), play a crucial role in several applications including optimizing smart grid systems but also introduce privacy risks. Our study analyzes the ability of forecasting models to leak global properties and privacy threats in smart grid systems. We demonstrate that a black box access to an LSTM model can reveal a significant amount of information equivalent to having access to the data itself (with the difference being as low as 1% in Area Under the ROC Curve). This highlights the importance of protecting forecasting models at the same level as the data.

Aim: Cardiac autonomic neuropathy (CAN) has been suggested to be associated with hypoglycemia and impaired hypoglycemia unawareness. We have assessed the relationship between CAN and extensive measures of glucose variability (GV) in patients with type 1 and type 2 diabetes. Methods: Participants with diabetes underwent continuous glucose monitoring (CGM) to obtain measures of GV and the extent of hyperglycemia and hypoglycemia and cardiovascular autonomic reflex testing. Results: Of the 40 participants (20 T1DM and 20 T2DM) (aged 40.70 +/- 13.73 years, diabetes duration 14.43 +/- 7.35 years, HbA1c 8.85 +/- 1.70%), 23 (57.5%) had CAN. Despite a lower coefficient of variation (CV) (31.26 +/- 11.87 vs. 40.33 +/- 11.03, P = 0.018), they had a higher CONGA (8.42 +/- 2.58 vs. 6.68 +/- 1.88, P = 0.024) with a lower median LBGI (1.60 (range: 0.20-3.50) vs. 4.90 (range: 3.20-7.40), P = 0.010) and percentage median time spent in hypoglycemia (4 (range:4-13) vs. 1 (range:0-5), P = 0.008), compared to those without CAN. The percentage GRADEEuglycemia (3.30 +/- 2.78 vs. 5.69 +/- 3.09, P = 0.017) and GRADEHypoglycemia (0.3 (range: 0 - 3.80) vs. 1.8 (range: 0.9-6.5), P = 0.036) were significantly lower, while the percentage median GRADEHyper-glycemia (95.45 (range:93-98) vs. 91.6 (82.8-95.1), P = 0.013) was significantly higher in participants with CAN compared to those without CAN. Conclusion: CAN was associated with increased glycemic variability with less time in euglycemia attributed to a greater time in hyperglycemia but not hypoglycemia.

The ransomware threat has loomed over our digital life since 1989. Criminals use this type of cyber attack to lock or encrypt victims' data, often coercing them to pay exorbitant amounts in ransom. The damage ransomware causes ranges from monetary losses paid for ransom at best to endangering human lives. Cryptographic ransomware, where attackers encrypt the victim's data, stands as the predominant ransomware variant. The primary characteristics of these attacks have remained the same since the first ransomware attack. For this reason, we consider this a key factor differentiating ransomware from other cyber attacks, making it vital in tackling the threat of cryptographic ransomware. This paper proposes a cyber kill chain that describes the modern crypto-ransomware attack. The survey focuses on the Encryption phase as described in our proposed cyber kill chain and its detection techniques. We identify three main methods used in detecting encryption-related activities by ransomware, namely API and System calls, I/O monitoring, and file system activities monitoring. Machine learning (ML) is a tool used in all three identified methodologies, and some of the issues within the ML domain related to this survey are also covered as part of their respective methodologies. The survey of selected proposals is conducted through the prism of those three methodologies, showcasing the importance of detecting ransomware during pre-encryption and encryption activities and the windows of opportunity to do so. We also examine commercial crypto-ransomware protection and detection offerings and show the gap between academic research and commercial applications.

Audio quality plays an essential role in several applications ranging from music to voice conversations. Sound information is subject to quality loss caused by reasons such as intermittent network connections, or storage corruption. Recent approaches resorted to using GANs for audio reconstruction due to their successful deployment in visual applications. However, audio datasets often include sounds from different contexts which increase the complexity of the patterns to be learned, leading to sub-optimal quality reconstruction. We propose a novel audio completion pipeline that clusters audio based on similarity of features extracted by a pre-trained CNN model and then trains a dedicated specialized GAN for each context separately. The proposed technique is compared with the traditional method of training one general GAN in completing 200ms missing segments of 1-second audio samples. Experimental results on a public benchmark dataset show that using specialized GANs led to a clear improvement in the completion quality as measured by a higher PSNR and a lower MSE. Qualitative evaluation also supported these results.

Detection of electricity theft improves the sustainability of the smart grid, helps electrical utilities mitigate their financial risks, and improves the overall management of resources. In this work, we utilize an LSTM neural network to forecast a given day’s energy consumption and construct residuals. The residuals are then compared to previous residuals from normal days using the Wasserstein distance. If the Wasserstein distance for the residuals of a day exceeds a threshold, the day is highlighted to indicate suspected energy theft. Our framework can be built upon existing forecasting models with minimal computational overhead to calculate the Wasserstein distance. The framework is also highly explainable, which reduces the cost of false positives significantly. Our framework was evaluated using a public dataset and was able to detect six attack models of energy theft and faulty meters, with a false positive rate of 9% and an average F1 score of 0.91.

Glucose monitoring is key to the management of diabetes mellitus to maintain optimal glucose control whilst avoiding hypoglycemia. Non-invasive continuous glucose monitoring techniques have evolved considerably to replace finger prick testing, but still require sensor insertion. Physiological variables, such as heart rate and pulse pressure, change with blood glucose, especially during hypoglycemia, and could be used to predict hypoglycemia. To validate this approach, clinical studies that contemporaneously acquire physiological and continuous glucose variables are required. In this work, we provide insights from a clinical study undertaken to study the relationship between physiological variables obtained from a number of wearables and glucose levels. The clinical study included three screening tests to assess neuropathy and acquired data using wearable devices from 60 participants for four days. We highlight the challenges and provide recommendations to mitigate issues that may impact the validity of data capture to enable a valid interpretation of the outcomes.

PurposeDigital tools have been used to document cultural heritage with high-quality imaging and metadata. However, some of the historical assets are totally or partially unlabeled and some are physically damaged, which decreases their attractiveness and induces loss of value. This paper introduces a new framework that aims at tackling the cultural data enrichment challenge using machine learning.Design/methodology/approachThis framework focuses on the automatic annotation and metadata completion through new deep learning classification and annotation methods. It also addresses issues related to physically damaged heritage objects through a new image reconstruction approach based on supervised and unsupervised learning.FindingsThe authors evaluate approaches on a data set of cultural objects collected from various cultural institutions around the world. For annotation and classification part of this study, the authors proposed and implemented a hierarchical multimodal classifier that improves the quality of annotation and increases the accuracy of the model, thanks to the introduction of multitask multimodal learning. Regarding cultural data visual reconstruction, the proposed clustering-based method, which combines supervised and unsupervised learning is found to yield better quality completion than existing inpainting frameworks.Originality/valueThis research work is original in sense that it proposes new approaches for the cultural data enrichment, and to the authors’ knowledge, none of the existing enrichment approaches focus on providing an integrated framework based on machine learning to solve current challenges in cultural heritage. These challenges, which are identified by the authors are related to metadata annotation and visual reconstruction.

Over the past years, notable progress has been made towards fighting misinformation spread over social media, encouraging the development of many fact-checking systems. However, systems that operate over Arabic content are scarce. In this work, we bridge this gap by proposing Tahaqqaq (Verify), an Arabic real-time system that helps users verify claims over Twitter with several functionalities, such as identifying check-worthy claims, estimating credibility of users in terms of spreading fake news, and finding authoritative accounts. Tahaqqaq has a friendly online Web interface that supports various real-time user scenarios. In the same breath, we enable public access to Tahaqqaq services through a handy RESTful API. Finally, in terms of performance, multiple components of Tahaqqaq outperform the state-of-the-art models on Arabic datasets.

Human breath analysis of volatile organic compounds has gained significant attention recently because of its rapid and noninvasive potential to detect various metabolic diseases. The detection of ketones in the breath and blood is key to diagnosing and managing diabetic ketoacidosis (DKA) in patients with type 1 diabetes. It may also be of increasing importance to detect euglycemic ketoacidosis in patients with type 1 or type 2 diabetes or heart failure, treated with sodium-glucose transporter-2 inhibitors (SGLT2-i). The present research evaluates the efficiency of colorimetry for detecting acetone and ethanol in exhaled human breath with the response time, pH effect, temperature effect, concentration effect, and selectivity of dyes. Using the proposed multidye system, we obtained a detection limit of 0.0217 ppm for acetone and 0.029 ppm for ethanol in the detection range of 0.05-50 ppm. A smartphone-assisted unit consisting of a portable colorimetric device was used to detect relative red/green/blue values within 60 s of the interface for practical and real-time application. The developed method could be used for rapid, low-cost detection of ketones in patients with type 1 diabetes and DKA and patients with type 1 or type 2 diabetes or heart failure treated with SGLT2-I and euglycemic ketoacidosis.

With the proliferation of fake news in the last few years, especially during the COVID-19 period, combating the spread of misinformation has become an urgent need. Although automated fact-checking systems were proposed recently, they leave much to be desired in terms of accuracy and explainability. Therefore, involving humans during verification could make the process much easier and more reliable. In this work, we propose an automated approach to detect claims that have been already manually-verified by professional fact-checkers. Our proposed approach uses recent powerful BERT variants as point-wise rerankers. Additionally, we study the impact of using different fields of the verified claim during training and inference phases. Experimental results show that our proposed pipeline outperforms the state-of-the-art approaches on two English and one Arabic datasets.