Biblio

The growing diffusion of robotics in our daily life demands a deeper understanding of the mechanisms of trust in human-robot interaction. The performance of a robot is one of the most important factors influencing the trust of a human user. However, it is still unclear whether the circumstances in which a robot fails to affect the user's trust. We investigate how the perception of robot failures may influence the willingness of people to cooperate with the robot by following its instructions in a time-critical task. We conducted an experiment in which participants interacted with a robot that had previously failed in a related or an unrelated task. We hypothesized that users' observed and self-reported trust ratings would be higher in the condition where the robot has previously failed in an unrelated task. A proof-of-concept study with nine participants timidly confirms our hypothesis. At the same time, our results reveal some flaws in the design experimental, and encourage a future large scale study.

Text-based CAPTCHAs are still commonly used to attempt to prevent automated access to web services. By displaying an image of distorted text, they attempt to create a challenge image that OCR software can not interpret correctly, but a human user can easily determine the correct response to. This work focuses on a CAPTCHA used by a popular Chinese language question-and-answer website and how resilient it is to modern machine learning methods. While the majority of text-based CAPTCHAs focus on transcription tasks, the CAPTCHA solved in this work is based on localization of inverted symbols in a distorted image. A convolutional neural network (CNN) was created to evaluate the likelihood of a region in the image belonging to an inverted character. It is used with a feature map and clustering to identify potential locations of inverted characters. Training of the CNN was performed using curriculum learning and compared to other potential training methods. The proposed method was able to determine the correct response in 95.2% of cases of a simulated CAPTCHA and 67.6% on a set of real CAPTCHAs. Potential methods to increase difficulty of the CAPTCHA and the success rate of the automated solver are considered.

Given the growing sophistication of cyber attacks, designing a perfectly secure system is not generally possible. Quantitative security metrics are thus needed to measure and compare the relative security of proposed security designs and policies. Since the investigation of security breaches has shown a strong impact of human errors, ignoring the human user in computing these metrics can lead to misleading results. Despite this, and although security researchers have long observed the impact of human behavior on system security, few improvements have been made in designing systems that are resilient to the uncertainties in how humans interact with a cyber system. In this work, we develop an approach for including models of user behavior, emanating from the fields of social sciences and psychology, in the modeling of systems intended to be secure. We then illustrate how one of these models, namely general deterrence theory, can be used to study the effectiveness of the password security requirements policy and the frequency of security audits in a typical organization. Finally, we discuss the many challenges that arise when adopting such a modeling approach, and then present our recommendations for future work.