Users are confronted with a variety of different machine learning applications in many domains. To make this possible especially for applications relying on sensitive data, companies and developers are implementing Privacy Preserving Machine Learning (PPML) techniques what is already a challenge in itself. This study provides the first step for answering the question how to include the user’s preferences for a PPML technique into the privacy by design process, when developing a new application. The goal is to support developers and AI service providers when choosing a PPML technique that best reflects the users’ preferences. Based on discussions with privacy and PPML experts, we derived a framework that maps the characteristics of PPML to user acceptance criteria.

With the rising popularity of artificial intelligence (AI) applications, the use of the underlying models has spread to the general public. These AI models have limitations and biases, and knowing about their characteristics could promote their safe use. Although there is some information about AI models available, in the form of AI Model Cards, there is little research on how useful this information is for non-expert users. In this paper, we conduct an experiment to evaluate the effectiveness and perception of information quality of the Model Card of a currently available AI and compare it with shorter versions. The results show that participants can use the Model Card to answer questions about the AI, but they are less confident about their answers compared to shorter versions. In addition, the full Model Card is considered less understandable and interpretable compared with a short version. On the other hand, a short version had a negative effect on perceived trustworthiness of the AI, but in all cases the participants had a positive attitude towards seeking information about the AI.

Today, vast amounts of private and sensitive data are being shared across a variety of on-line services day-today. Recent technologies increasingly simplify the collection, processing and evaluation of these data. This results in numerous threats to the privacy of users. Although there are legal regulations to protect privacy, users are increasingly faced with the challenge of controlling their data to exercise their rights. In order to implement the existing legal framework and to help users protect their privacy, technological solutions are becoming increasingly important. Within the scope of this paper, the research areas of privacy risk detection will be examined in more detail. For this purpose, the state of the art of privacy sensitive information detection is elaborated and then analyzed by means of a specifically developed classification scheme to identify research gaps and trends. As a result, several research gaps and trends have been identified, demonstrating that further research is required to develop user tailored privacy enhancing tools and ensure adequate privacy protection.

Users report that they have regretted accidentally sharing personal information on social media. There have been proposals to help protect the privacy of these users, by providing tools which analyze text or images and detect personal information or privacy disclosure with the objective to alert the user of a privacy risk and transform the content. However, these proposals rely on having access to users’ data and users have reported that they have privacy concerns about the tools themselves. In this study, we investigate whether these privacy concerns are unique to privacy tools or whether they are comparable to privacy concerns about non-privacy tools that also process personal information. We conduct a user experiment to compare the level of privacy concern towards privacy tools and nonprivacy tools for text and image content, qualitatively analyze the reason for those privacy concerns, and evaluate which assurances are perceived to reduce that concern. The results show privacy tools are at a disadvantage: participants have a higher level of privacy concern about being surveilled by the privacy tools, and the same level concern about intrusion and secondary use of their personal information compared to non-privacy tools. In addition, the reasons for these concerns and assurances that are perceived to reduce privacy concern are also similar. We discuss what these results mean for the development of privacy tools that process user content.

Adversarial images have been proposed to deceive deep neural networks (DNNs) by adding perturbations to the pixels. Unlike existing attacks, Su et al. [1] analyzed an attack in an extremely limited constraint where only one pixel was modified. However, their one-pixel attack is easy to recognize by humans. In this paper, we improve the attack to enable the deceit of both DNNs and humans. We conducted a human recognition analysis to prove our attack's effect. We then propose detection and defense methods against the attack by re-attacking the adversarial images. Our experimental results on the six most recent convolutional neural networks show that while our attack achieved approximately the same success rates and confidence scores as in the existing attack, our attack achieves a higher success rate for deceiving humans. Only 49.41 % of participants can recognize our attack even though 81.04 % participants have recognized the existing attack. OPA2D detects 99.33% of the existing attack and 100% of our attack and defends 92.00% of the existing attack and 95.33 % of our attack.

In this paper, we consider the issue of trust and trust-related factors in the context of decentralized applications running on public blockchain platforms such as Ethereum. These decentralized applications emphasize a lack of reliance on a trusted third party, and are marketed as applications that cannot be censored or stopped. To determine whether either social trust or technology trust applies in these cases, we examine the extent to which these applications could be considered to be out of the control of a third party, by qualitatively analyzing how developers define the characteristics of decentralization, trustlessness and autonomy. The results show that although decentralized applications' websites make reference to these concepts, they are not defined in the same way. In cases where there is no mention of either of these concepts, it is therefore difficult to say which definitions are assumed. In addition, we also found contradictions in the characterization of the level of developer control. We discuss these findings in the context of research on user trust and propose future research directions.

Positive feedback from previous users can affect consumer behavior towards an electronic commerce website. Social Network Sites provide a way to gather feedback from SNS users and show them on websites, but not much research has been done on this type of feedback mechanism. In this study we investigate the influence that SNS-based feedback has on the behavior of consumers, comparing the effect in consumers who are members of a SNS vs. those who are not members. We conducted a survey in Japan, using a mock Thai website that showed three different levels of SNS information. We found that consumers who are SNS members have a more positive attitude towards the website when SNS-based feedback information is shown. We also found that the nationality of the SNS members giving the feedback affects trust and risk perceptions differently.