Mike Higgins

Joydeep Mitra, Venkatesh-Prasad Ranganath, Torben Amtoft et al.

Mobile apps provide various critical services, such as banking, communication, and healthcare. To this end, they have access to our personal information and have the ability to perform actions on our behalf. Hence, securing mobile apps is crucial to ensuring the privacy and safety of its users. Recent research efforts have focused on developing solutions to secure mobile ecosystems (i.e., app platforms, apps, and app stores), specifically in the context of detecting vulnerabilities in Android apps. Despite this attention, known vulnerabilities are often found in mobile apps, which can be exploited by malicious apps to harm the user. Further, fixing vulnerabilities after developing an app has downsides in terms of time, resources, user inconvenience, and information loss. In an attempt to address this concern, we have developed SeMA, a mobile app development methodology that builds on existing mobile app design artifacts such as storyboards. With SeMA, security is a first-class citizen in an app's design -- app designers and developers can collaborate to specify and reason about the security properties of an app at an abstract level without being distracted by implementation level details. Our realization of SeMA using Android Studio tooling demonstrates the methodology is complementary to existing design and development practices. An evaluation of the effectiveness of SeMA shows the methodology can detect and help prevent 49 vulnerabilities known to occur in Android apps. Further, a usability study of the methodology involving ten real-world developers shows the methodology is likely to reduce the development time and help developers uncover and prevent known vulnerabilities while designing apps.

Matthew Dunn, Levent Sagun, Mike Higgins et al.

We publicly release a new large-scale dataset, called SearchQA, for machine comprehension, or question-answering. Unlike recently released datasets, such as DeepMind CNN/DailyMail and SQuAD, the proposed SearchQA was constructed to reflect a full pipeline of general question-answering. That is, we start not from an existing article and generate a question-answer pair, but start from an existing question-answer pair, crawled from J! Archive, and augment it with text snippets retrieved by Google. Following this approach, we built SearchQA, which consists of more than 140k question-answer pairs with each pair having 49.6 snippets on average. Each question-answer-context tuple of the SearchQA comes with additional meta-data such as the snippet's URL, which we believe will be valuable resources for future research. We conduct human evaluation as well as test two baseline methods, one simple word selection and the other deep learning based, on the SearchQA. We show that there is a meaningful gap between the human and machine performances. This suggests that the proposed dataset could well serve as a benchmark for question-answering.