SkillBot: Identifying Risky Content for Children in Alexa Skills

As smart speakers such as the Amazon Echo have grown in popularity, these devices have presented new opportunities for exploring voice interaction, particularly in the domain of exercise. We have created TandemTrack, a multimodal system comprised of a mobile app and an Alexa skill whose capabilities include exercise guidance, data capture, feedback, and reminder. We propose using TandemTrack in a deployment study to explore and analyze individuals' preferences and use of voice and visual interactions, and to test TandemTrack's effectiveness as a whole in promoting consistent exercise habits while additionally comparing the effectiveness of the app against the skill. Not only does TandemTrack make important steps in addressing gaps in currently available exercise assistants, but also, more importantly, we believe this is the first study to investigate the specific value that voice interactions may provide over other technologies in promoting consistent exercise.

Third-party applications frequently request access to various types of user data, such as contacts, photos, and location, to provide enhanced functionality and improve user experience. For instance, social media platforms may integrate with third-party apps that facilitate post scheduling, engagement analytics, or introduce additional features not available on the platform itself. However, enabling these capabilities often requires granting third-party apps permission to access sensitive data, which introduces potential privacy and security concerns. Once access is approved, these applications can collect, use, and occasionally share user data with other entities, increasing the risk of privacy violations if data protection measures are inadequate or the data is misused. Granting such permissions may expose users to security vulnerabilities, particularly if the third-party apps lack robust security frameworks, making user data susceptible to hacking or unauthorized access.

Employers' selection practices sometimes involve reviewing applicants' profile on social networking websites (SNWs) and invading applicants' privacy (e.g., asking for their passwords). Applicants can be eliminated because of faux pas (i.e., inappropriate content) they post online. Yet, little research has examined factors related to faux pas postings. The present study examines employers' use of SNWs in selection, participants' internet and SNWs use, personality, and SNWs self‐promotion as predictors of the likelihood of faux pas postings. Results show lower likelihood of faux pas postings when participants are informed that a high proportion of employers use SNWs in selection, but mainly when it includes invasion of applicants' privacy. Moreover, participants' age, privacy settings, extraversion, and SNWs self‐promotion are related to faux pas.

Signal recognition is vital for sectors like medical diagnosis, security monitoring, intelligent transportation, and voice interaction. Traditional methods, however, relying on manually designed features, struggle with complex patterns and high-dimensional signals. While machine learningespecially deep learningaddresses these issues via end-to-end learning and automatic feature representation, it suffers from over-reliance on high-quality labeled data. Real-world challenges such as signal noise and high labeling costs lead to label noise, impeding practical application. This review explores signal recognition fundamentals, including feature extraction, selection and classification. It details deep learning applications in image recognition, like CNNs, RNNs and Transformers. It also discusses multimodal learning, using AV-ASR and BPO-AVASR as examples. Finally, it identifies challenges (data scarcity, high model complexity, privacy issues) and proposes future directions (lightweight models, few-shot learning). The review concludes that deep learning dominates signal recognition, with models achieving human-level performance on benchmarks. Multimodal learning, fusing speech and image data, is a key trend.

With 20 million installs a day [1], third-party apps are a major reason for the popularity and addictiveness of Facebook. Unfortunately, hackers have realized the potential of using apps for spreading malware and spam. The problem is already significant, as we find that at least 13% of apps in our dataset are malicious. So far, the research community has focused on detecting malicious posts and campaigns. In this paper, we ask the question: Given a Facebook application, can we determine if it is malicious? Our key contribution is in developing FRAppE---Facebook's Rigorous Application Evaluator---arguably the first tool focused on detecting malicious apps on Facebook. To develop FRAppE, we use information gathered by observing the posting behavior of 111K Facebook apps seen across 2.2 million users on Facebook. First, we identify a set of features that help us distinguish malicious apps from benign ones. For example, we find that malicious apps often share names with other apps, and they typically request fewer permissions than benign apps. Second, leveraging these distinguishing features, we show that FRAppE can detect malicious apps with 99.5% accuracy, with no false positives and a high true positive rate (95.9%). Finally, we explore the ecosystem of malicious Facebook apps and identify mechanisms that these apps use to propagate. Interestingly, we find that many apps collude and support each other; in our dataset, we find 1584 apps enabling the viral propagation of 3723 other apps through their posts. Long term, we see FRAppE as a step toward creating an independent watchdog for app assessment and ranking, so as to warn Facebook users before installing apps.

The goal of this study is to perform an analysis of natural language models of interaction, associated to 3 types of input devices with a microphone, a hands-free device, a mobile app and a remote control, in order to identify the most suited to be adopted in a conversational voice solution for interactive television in Portuguese. The research addressed issues associated with natural language systems such as usability, interaction and privacy, using a prototype based on a Wizard-of-Oz methodology. It was possible to verify that the interaction model that met the majority of preferences was the hands-free solution, based on a far-field microphone activated by a wake-up word. Aside from easy to use, it was the one that raised the least difficulties of execution. Despite this, the interaction model based on the remote control with microphone was not completely set aside. Some participants pointed out that the remote control, as a device inherently associated with television, is the most natural way of interaction. Although most participants indicate that a natural language interaction system does not raise significant privacy issues, it is important to have this aspect in consideration when choosing the interaction model to be adopted on a voice system commercial solution. To top it off, participants consider that a voice-operated system, like the one they were evaluating, is very useful and almost all have been receptive to having such a system at home.

User's Awareness of Privacy Issues Presented by Third Party Applications on Facebook

Android devices are popular nowadays due to their rich functionality and good performance. Android provides a permission model that supports user privacy by protecting restricted data access and actions on the Android operating system. Based on the app's defined functionality, any app must get users' permission to access different types of restricted data or resources. However, some developers request additional permissions other than the app's functionality, known as over-claim permissions. These over-claim permissions may steal sensitive information from users, Which is one of the leading privacy issues. Looking towards its criticality, many researchers have started working on detecting over-claim permission. This paper conducts a comprehensive survey of various approaches to identify and handle over-claim permissions. It also discusses the limitations of existing techniques and further research scope in this direction, which may help researchers define a path for designing new methods or optimizing existing strategies to resolve over-claim permission problems satisfactorily.

Focal Area(s): 3. Insight gleaned from complex data (both observed and simulated) using AI, big data analytics, and other advanced methods, including explainable AI and physics- or knowledge-guided AI; & 1. Data acquisition and assimilation enabled by machine learning, AI, and advanced methods including experimental/network design/optimization, unsupervised learning (including deep learning), and hardware-related efforts involving AI (e.g., edge computing). Science Challenge: Interactions between water, land, and energy systems are complex and occur on a variety of scales, ranging from local to basinal to regional. Accurately predicting the behavior of ground water and surface water systems for 5-10 years and beyond requires an understanding of the current system and the ability to model both the natural system at scale and human-induced forcings related to energy and other activities. Artificial intelligence and machine learning (AI/ML) combined with modern compilation and integration efforts for U.S. groundwater and surface water systems present potential solutions to bolstering detailed physics-based models of these systems. Big data tied with ML and physics-based modeling can drive breakthroughs in understanding the earth system, but research is often impeded by data access (e.g., privacy issues), quality, formats, gaps, multi-source, multi-scale, integration, and spatiotemporal challenges. Effective integration of real data and simulated (synthetic) data that fill gaps is critical. Overcoming these complex data and model integration challenges will enable a transformational approach to acquiring enhanced understanding of environmental systems.

Through improvements of voice recognition smart voice assistants have become available in various forms. They offer users the possibility to control devices by natural voice interaction. To improve accuracy, user conversation data is automatically streamed to companies for machine learning. This is perceived as an act privacy invasion by many users. This is also an argument for potential buyers against acquiring a voice assistant device. Therefore, the central design element of smart home voice assistants should be local processing. For this purpose, the work was based on the Privacy by Design principles by Cavoukian. Based on the analysis of existing cloud and offline solutions a prototype was developed. Through complete offline modality features like self adapting commands and more natural forms of voice interaction were integrated. This was possible while protecting user privacy. Furthermore, fulfilment of the goals with focus on accuracy and quick response time was evaluated. This paper shows that it is possible to implement an offline voice assistant with similar response times and high accuracy. Self adaptation and expanded recognition show great potential for future improvements.

Social media, a widely used communication platform, has significant implications, especially for children and adolescents.This literature review examines the unethical marketing practices in social media and their harmful effects on younger audiences.The study highlights the exposure of youth to inappropriate content, privacy invasions, and unhealthy dietary influences through targeted advertising.The review also discusses the ethical concerns surrounding the use of artificial intelligence in marketing.The paper calls for stricter regulations, further research, and enhanced transparency to mitigate these negative impacts.

SUMMARY Attempts to formulate appropriate policies for the management of agricultural systems are constrained by the limitations of acquiring information about changes in the natural system across space and through time. The adoption of natural science techniques to measure the transformation of natural phenomena requires considerable investment in time and expertise and may not always be capable of providing information about change in a way that meets the adaptive requirements of planning and policy. A more flexible approach to data collection will require the acceptance of less precise and more general physical data as the basis for understanding changes in complex agricultural systems. This raises interesting questions about the adoption of social enquiry techniques to provide information about changing natural phenomena. The progressive deterioration of water resources, as a result of intensive agricultural activity in the Argolid Plain of Greece, will be considered and the problems of measuring the ...

The popularity of third-party apps on social network sites and mobile networks emphasizes the problem of the interdependency of privacy. It is caused by users installing apps that often collect and potentially misuse the personal information of users’ friends who are typically not involved in the decision-making process. In this paper, we provide an economic model and simulation results addressing this problem space. We study the adoption of social apps in a network where privacy consequences are interdependent. Motivated by research in behavioral economics, we extend the model to account for users’ other-regarding preferences; that is, users care about privacy harms they inflict on their peers.We present results from two simulations utilizing an underlying scale-free network topology to investigate users’ app adoption behaviors in both the initial adoption period and the late adoption phase. The first simulation predictably shows that in the early adoption period, app adoption rates will increase when (1) the interdependent privacy harm caused by an app is lower, (2) installation cost decreases, or (3) network size increases. Surprisingly, we find from the second simulation that app rankings frequently will not accurately reflect the level of interdependent privacy harm when simultaneously considering the adoption results of multiple apps. Given that in the late adoption phase, users make their installation decisions mainly based on app rankings, the simulation results demonstrate that even rational actors who consider their peers’ well-being might adopt apps with significant interdependent privacy harms. Our findings complement the usable privacy and security studies which show that users install privacy-invasive apps because they are unable to identify and understand apps’ privacy consequences; however, we show that fully-informed and rational users will likely fall for privacy-invasive apps as well.

Modern enterprises are actively implementing artificial intelligence and automated natural language processing systems to optimize work with corporate information. However, the use of these technologies carries significant risks of data privacy violations, especially when analyzing unstructured text materials. Existing cryptographic methods of information protection demonstrate limited effectiveness in processing linguistic data, and modern approaches to secure information processing have significant disadvantages: homomorphic encryption methods significantly reduce computing speed, and distributed learning technologies face problems with matching heterogeneous data. In the presented study, a comprehensive information security method has been developed that combines the advantages of homomorphic encryption and distributed learning. The main scientific achievement of the work is the creation of specialized cryptographic algorithms for processing language data, including the identification of named objects in encrypted form, as well as optimizing the ratio between the level of protection and computational efficiency. The proposed methodology includes differentiated encryption of various types of data, decentralized information processing with automatic adjustment of learning parameters, as well as mechanisms to ensure anonymity and access control. The experimental part of the study evaluates three key parameters: the system's resilience to potential attacks, the processing speed of protected data, and the accuracy of analytical algorithms. The practical value of the work lies in creating a secure platform for the implementation of intelligent data analysis systems in the financial sector, healthcare and telecommunications industries, where information leakage can lead to serious consequences. A promising area of further research is the development of cryptographic algorithms that are resistant to quantum hacking methods.

The NIST Privacy Framework describes itself as a comprehensive approach to organization-wide privacy program management. However, inferences can yield sensitive information of identities or attributes from nonsensitive information. Privacy governance must protect this information. Although many people and organizations are expanding their privacy definitions to include inferences, our gap analysis reveals that the framework's mapped controls are insufficient for managing inference-driven risk. The framework does not attend organizational focus to privacy inference risk sufficiently to support its stated claim of comprehensive risk management. Applying the framework to past incidents where ostensibly protected information was re-inferred, we analyze how organizations can better mitigate inference-based privacy violations. Finally, we recommend detailed improvements to the framework's controls to account better for inferences. Our recommendations encompass augmenting and mapping additional privacy risk controls to increase implementing organizations' awareness of inference risks, updating controls that depend on protecting specific PII categories, and enhancing organizations' proficiency in translating legal and policy requirements into technical implementations.