ABSTRACT. Breast cancer is reported as one of most common malignancy amongst women in the world. Early detection of this cancer is critical to clinical and epidemiologic for aiding in informing subsequent treatments. This study investigates automated breast cancer prediction using deep learning techniques. A new 19-layer deep convolutional neural network (CNN) model for detecting the benign breast tumors from malignant cancers was proposed and implemented. The experiments on BreaKHis dataset was conducted and K-fold Cross Validation technique are used for the model evaluation. The proposed 19-layer deep CNN based classifiers compared with conventional machine learning classifier, namely Support Vector Machine (SVM) and a state-of-the-art deep learning model, namely GoogLeNet in terms of Accuracy, Area under the Receiver Operating Characteristic (ROC) Curve (AUC), the Classification Mean Absolute Error (MAE), Mean Squared Error (MSE) metrics. The results demonstrate that the proposed new model outperformed the other classifiers. The proposed model achieved an accuracy, AUC, MAE and MSE of 84.5%, 85.7%, 0.082, and 0.043, respectively.

ABSTRACT. Stakeholders in the insurance industry are committed to yet lack the timely and actionable information for alleviating policyholder's mental health concerns and the industry's mental health climate. Existing research has revealed that personal data, such as depression, anxiety, and stress, can provide deeper insights into policyholder's mental health states. However, such data remain unexplored for supporting stakeholders' and government goals. In this paper, we design an interactive visualization system to provide deeper insight into policyholder's mental health states and performs recommendation reasoning. Our study has three implications: (i) insurance data are potentially useful for understanding policyholder's mental health; (ii) a dashboard-like visual representation is helpful for the decision-making of Stakeholders; and (iii) recommendations on how the government can improve the mental health of Australians. We conclude by reporting an informal evaluation of the effectiveness of our system and remarking on the future directions in system design.

ABSTRACT. Electricity Market uses Demand and Supply chain strategy. Also, it is prone to random fluctuations that directly impact profit. Therefore forecasting demand becomes very important to mitigate the consequences of price dynamics. The paper proposes the Deep Learning model using Long Short Term Memory (LSTM) and Convolution Neural Network to forecast future electricity prices on the Australian electricity market and compares them with other state of the art models. We have selected evaluation metrics to prove that our model outperforms the other existing models for electricity price prediction.

ABSTRACT. Analysis of the human mobility from GPS trajectories becomes crucial in many aspects such as policy planning for urban citizens, location-based service recommendation/prediction, and especially mitigating the spread of biological and mobile virus. In this paper, we propose a method to find which temporal factors affecting human mobility lifestyle. We collected GPS data from 100 smartphone users and designed a multiple linear regression model which consists of 13 temporal patterns. Our result found that people tend to keep their mobility habits on Thursday and on the days in the second week of a month but tend to lose the habits on Friday. We also gave an explanation behind these findings.

ABSTRACT. The prosocial behavior plays an important role in the containing of COVID-19 pandemic. The factors that could impact prosocial behavior and its facilitation mechanisms need further investigation. In this study, the effects of individuals' national identity and subjective perceptions of the COVID-19 pandemic on prosocial behavior were explored. From February to March 2020, 256 questionnaires were obtained. The national identity, prosocial behavior, and perceptions of the degrees of severity, scarcity of resources, controllability, and familiarity of the pandemic were measured. It is found that the prosocial behavior increases with national identity. The perception of the degree of severity of the pandemic plays a moderating role in the relationship between the national identity and prosocial behavior. To the ingroup prosocial behavior, there is no significant interaction between the national identity and the degree of perceived severity. Nevertheless, the outgroup prosocial behavior was more impacted by national identity when the perception of the degree of severity was relatively low. Additionally, the perception of the degree of controllability plays a mediating role in the relationship between the national identity and prosocial behavior (especially to the outgroup). In conclusion, the national identity and subjective perceptions of the COVID-19 epidemic affect the prosocial behavior, but with different impact mechanisms on ingroup and outgroup members. To accomplish the great success in combating the COVID-19 pandemic by promoting prosocial behavior, the society, government, and individuals should facilitate the national identity and advance the understanding of the epidemic.

ABSTRACT. With the great development of 5G technology research and the constant development of the network shopping, clothing classification and clothing collocation recommendation based on clothing pictures can provide advices to the customer and help businesses to promote sales. Deep learning is a latest research achievement in the field of machine learning, it has a strong ability of image modeling and image representation, which makes breakthrough progress in the field of image processing. Based on image data of dressing commodities provided by Taobao.com, as well as the text data of both users' historical behaviors and dressing outfits generated by fashion experts, we design and implement clothing collocation and recommendation through relevant technologies of data mining and deep learning.

ABSTRACT. In this paper, we analyze over 18 million Coronavirus related Twitter discussions (tweets) collected between Mar 1, 2020 and May 31, 2020. We perform sentiment analysis to evaluate the relationship between the the public’s sentiments and the number of COVID-19 cases. We also look at the frequency of mentions of a country in tweets and the rise in its’ daily number of COVID-19 cases. Some of our findings include the discovery of a correlation between the number of tweets mentioning Italy, USA, and UK and the daily increase in new COVID-19 cases in these countries.

ABSTRACT. Much research has taken place aiming to understand the role of in-game behavior, particularly, moral behaviors in video games. However, less research has examined the design of these moral decisions and how it could influence the in-game and real-life decision-making process, such as meta-choices. Meta-choices are the choices above that of the game itself, for example the choice to stop playing the game. This research aimed to understand in-game moral behavior with restricted options in the game. Participants (N = 115) played a purpose made game where only anti-social options were presented as an in-game choice to examine if a meta-choice would be made. It was found that eight participants considered stopping the game and only two participants made the meta-choice to stop playing. Overall, this suggests a potential influence and bias in decision-making; the presented choice would be selected rather than the meta-choice to stop playing.

ABSTRACT. In this paper we evaluate the sentiment of messages by Russian Internet Research Agency (IRA) on Twitter discourse during US Presidential Elections of 2016 using VADER. We use two datasets for analysis. The first consists of 51.3 million tweets collected during the US Elections of 2016 (October 30th, 2016–November 18th, 2016) and the second was shared by Twitter in October 2018, consisting of 8.77 million tweets generated by IRA accounts over a decade. We look for overlap of IRA tweets in the two datasets, evaluate their sentiment, and compare it with sentiment of other messages during that time period discussing the two Presidential candidates. Our findings show: (1) IRA tweets and retweets had a significantly positive sentiment towards Donald Trump and negative sentiment towards Hillary Clinton; (2) IRA messages mentioning Hillary Clinton had a more negative sentiment than non-IRA messages in our dataset.

ABSTRACT. Contact tracing apps are powerful software tools that can help control the spread of COVID-19. In this article, we evaluated 53 COVID-19 contact tracing apps found on the Google Play Store in terms of their usage, rating, access permission, and user privacy. For each app included in the study, we identified the country of origin, number of downloads, and access permissions to further understand the attributes and ratings of the apps. Our results show that contact tracing apps had low overall privacy ratings and nearly 40% of the included apps were requesting “dangerous access permission” including access to storage, media files, and camera permissions. We also found that user adoption rates were inversely correlated to access permission requirements. To the best of our knowledge, our article summarizes the most extensive collection of contact tracing apps for COVID-19. We recommend that future contact tracing apps should be more transparent in permission requirements and should provide justification for permissions requested to preserve the app users’ privacy.

ABSTRACT. Digital Identity has the similar importance in the digital world to represent us just as our physical identity in the real world. Digital identity comprises certain personal and confidential attributes related to us and managed through an identity management system (IDM). In most IDM systems, identity owners do not control identity and its related personal data. However, self-sovereign identity (SSI) is an emerging IDM system which offers the ownership and full control of our personal data. General Data Protection Regulation (GDPR) is the basic requirement for anyone who deals with the personal data and SSI requires handling of identity and its associated personal data. If an SSI system comply with GDPR principles then it could become the most suitable IDM solution legally and universally. This paper evaluates this aspect of SSI and analyses SSI compliance with the key principles of GDPR. It also evaluates two different types of SSI ecosystems uPort and Sovrin based on the various roles and their compatibility with GDPR roles. Finally, this paper performs their comparative analysis to check their compliance with the key principles of GDPR.

ABSTRACT. Contact Tracing (CT) is seen as a key tool in reducing the propagation of viruses, such as Covid-19. Given near ubiquitous societal usage of mobile devices, governments globally are choosing to augment manual CT with CT applications (CTAs) on smart phones. While a plethora of solutions have been spawned, their overall effectiveness is based on majority population uptake. Unfortunately, their rapid deployment and the nature of the information they gather has prompted a variety of user concerns such as information privacy and Data Protection (DP). Therefore selecting an optimal solution to maximise user trust and uptake is crucial. In this work, we present our initial CTA evaluation taxonomy for societal concerns. This is a subset of our larger taxonomy which is being developed as part of the Science Foundation Ireland project - COVIGILANT, the goal of which is to evaluate and compare numerous CTAs to select the optimal solution for a given population. In this paper we present a novel approach for evaluating CTAs with respect to the societal concerns of security, data protection and transparency. We then highlight the resultant challenges to data protection for now and future CTAs.

ABSTRACT. Customers are the core of businesses. Specifically, telecommunication companies, customer satisfaction is considered to be a commercial necessity and therefore a priority. High rates of customer satisfaction increase both retention and attraction rates. As a result, telecommunication companies are always seeking new means to achieve these objectives. A large volume of calls is received in a typical call center from customers complaining about phishing or spam attacks daily. It is difficult to identify the purpose of the call manually. In this work, we expand on previous efforts to focus more on impacted phone spam or phish consumers. The study focuses on both mediums of communication, phone calls and messages. A historical sample of customers' complaints dataset was used, and several machine learning classification algorithms were applied to analyze the calls. These are Logistic Regression, XGBoost, Gradient Boosting, Random Forest, CatBoost, KNN, and SVM. The predictive model can identify whether an individual is likely complaining or to complain about a spam or phish attack. The performance of the baseline classifier achieves an accuracy of 63.4\% that is based on CatBoost. Moreover, the model identifies consumers' demographics. The findings show that people of age 45 are more likely to complain and that males are less likely to complain.

ABSTRACT. We propose a novel conditional GANs architecture for crime (robberies) prediction in Bogotá, capital city of Colombia. The model uses several layers of ConvLSTM neural nets in both the generative and the discriminatory networks. We further condition on past crime intensity maps, weekdays, and holidays. The trained network is able to capture spatiotemporal patterns and outperforms state-of-the-art predictive models such as spatiotemporal Poisson point process, as well as other models trained with the same dataset. Model’s accuracy reaches an area under the Hit Rate – Percentage Area Covered by Hotspots curve of 0.86. However, our predictions suggest that there is a potential bias with heterogeneous effects on vulnerable populations. We address the fairness consequence of this model in low income vs. high income residents by estimating a calibration test conditional to these protected variables. Finally, we introduce a fairness – accuracy balancing technique that quantifies the tradeoffs between accuracy and fairness in this type of models. This technique notably reduces bias with a marginal effect on accuracy.

ABSTRACT. A fundamental aspect of the perception of security is the fear of crime, which is the concern of being a crime victim. The fear of crime has negative social consequences, including neighborhood deterioration, physical and behavioral health outcomes, among others. Different interventions allow fear of reduction, including crime reduction, an increase of police presence, and improvement of social cohesion, among others. However, there are no quantitative approaches to guide the selection of policies for reducing the fear of crime. This article proposes a novel method based on optimization for finding policies aimed to decrease fear of crime by using mathematical models and evolutionary algorithms. Results suggest that policies that promote interactions among members of different groups may enhance community cohesion resulting in reductions of the fear of crime for the most susceptible members in the group.

ABSTRACT. The current consensus is that a lack of skilled young persons entering the cyber security industry is contributing significantly to the accrescent cyber security skills gap. However, little progress has been made in terms of handling key contributing factors – cyber security education. While Capture The Flag(CTF) exercises in cyber security education present some of the necessary requirements, we hypothesise that the current CTF forms do not possess the requirements necessary for promoting student engagement and learning. The paper presents the results of a study aimed at identifying the requirements of a student-focused CTF.

ABSTRACT. Aggressive behaviors are acts (through physical, verbal, or psychological means) that can cause harm, pain, or injury to another person. In Colombia, aggressive behaviors that shock public peace or that are reported to the emergency line are classified as quarrels. About a million quarrels were reported in Bogota city during 2017-2018, and 70% of these incidents generated personal injuries or homicides. Considering these statistics, the characterization and prediction of this phenomenon adopt relevance in the security agendas of decision-makers. The identification of temporal patterns in the occurrence of aggressive behavior might help to get a better understanding of the phenomenon and develop more accurate predictive models. In this paper, we propose a characterization of temporal patterns in the occurrence of aggressive behaviors in Bogota city. The characterization is developed through predictability quantification using Colwell’s predictability measures. Results suggest that aggressive acts are more predictable in some areas and that (in most cases) predictability is associated with the existence of cyclic or temporal patterns in the occurrence of this type of incident.

ABSTRACT. Predicting crime is extremely desired by police departments across the world. However, since society tends to be more reluctant to more violent and costly crimes such as homicide, not all types of crime have the same priority in the agendas of policymakers. Relative to other types of crime, homicides are statistically more challenging due to its sparsity and low frequency. For instance, over the last five years the average number of homicides across the city of Bogota has been roughly a thousand events per year, compared to the more than one hundred thousand robberies reported in the same period. Although, more than 80% of the homicides in the city occur during street fights suggesting a strong spatial and temporal correlation between these two types of crime. With this in mind, we used a manifold learning approach that capitalizes on a rich dataset of street fights to discover a criminal manifold that we use to penalize a KDE model of homicides where sparsity and low frequency is an issue. To implement this we follow a Kernel Warping methodology (Zhou & Matteson, 2015). The methodology reduces the relevant space for homicide prediction to regions of the city where homicides or street fights have occurred, giving more weight to the homicide episodes. We also introduce a temporal decay component to place a larger importance to recent events. The proposed model outperforms a standard KDE trained with homicide data, and a KDE trained in both homicide and street fights data for homicide prediction: flagging the 5% of the area of the city with the highest estimated density, the Kernel Warping model correctly identifies between 30% and 35% of the homicides in the test set.

ABSTRACT. Action Fraud reported 50 million pounds was lost to romance fraud in 2018, a 27 percent increase on the previous year, despite an increase in publicity and guidance surrounding the issue. Romance fraud is an ever-increasing issue, and the statistics highlight the need for a proactive, adaptable, and bespoke approach to assist online dating platforms in combatting the problem, providing targeted awareness to customers while improving the user experience of dating platforms. Currently, there is no effective approach for increasing user awareness and providing real-time intervention of romance fraud. Existing methods on the platform focus on identifying, preventing, and stopping threat actors with technological measures rather than educating potential victims. This paper discusses the existing state of romance fraud and proposes a solution to mitigate the problems by developing a targeted awareness approach. The solution can be adopted by online dating platforms for improving early identification and intervention. It includes bespoke advisory messages to be provided to the user and risk categorisation criteria as well as workflows and prototypes to assist platforms with implementation. The results from the primary research clearly support the objectives showing that early intervention helps to mitigate against fraud, decreasing the likelihood of it occurring. This approach offers demonstrable improvements to dating platforms.