Wearable devices, such as smartwatches, are furnished with state-of-the-art sensors that enable a range of context-aware applications. However, malicious applications can misuse these sensors, if access is left unaudited. In this paper, we demonstrate how applications that have access to motion or inertial sensor data on a modern smartwatch can recover text typed on an external QWERTY keyboard. Due to the distinct nature of the perceptible motion sensor data, earlier research efforts on emanation based keystroke inference attacks are not readily applicable in this scenario. The proposed novel attack framework characterizes wrist movements (captured by the inertial sensors of the smartwatch worn on the wrist) observed during typing, based on the relative physical position of keys and the direction of transition between pairs of keys. Eavesdropped keystroke characteristics are then matched to candidate words in a dictionary. Multiple evaluations show that our keystroke inference framework has an alarmingly high classification accuracy and word recovery rate. With the information recovered from the wrist movements perceptible by a smartwatch, we exemplify the risks associated with unaudited access to seemingly innocuous sensors (e.g., accelerometers and gyroscopes) of wearable devices. As part of our efforts towards preventing such side-channel attacks, we also develop and evaluate a novel context-aware protection framework which can be used to automatically disable (or downgrade) access to motion sensors, whenever typing activity is detected.

We propose a novel framework for measuring and evaluating location privacy preserving mechanisms in mobile wireless networks. Within this framework, we first present a formal model of the system, which provides an efficient representation of the network users, the adversaries, the location privacy preserving mechanisms and the resulting location privacy of the users. This model is general enough to accurately express and analyze a variety of location privacy metrics that were proposed earlier. By using the proposed model, we provide formal representations of four metrics among the most relevant categories of location privacy metrics. We also present a detailed comparative analysis of these metrics based on a set of criteria for location privacy measurement. Finally, we propose a novel and effective metric for measuring location privacy, called the distortion-based metric, which satisfies these criteria for privacy measurement and is capable of capturing the mobile users' location privacy more precisely than the existing metrics. Our metric estimates location privacy as the expected distortion in the reconstructed users' trajectories by an adversary.

Abstract-

Personal and contextual information are increasingly shared via mobile social networks. Users' locations, activities and their co-presence can be shared easily with online "friends", as their smartphones already access such information from embedded sensors and storage. Yet, people usually exhibit selective sharing behavior depending on contextual attributes, thus showing that privacy, utility, and usability are paramount to the success of such online services. In this paper, we present SPISM, a novel information-sharing system that decides (semi-)automatically whether to share information with others, whenever they request it, and at what granularity. Based on active machine learning and context, SPISM adapts to each user's behavior and it predicts the level of detail for each sharing decision, without revealing any personal information to a third-party. Based on a personalized survey about information sharing involving 70 participants, our results provide insight into the most influential features behind a sharing decision. Moreover, we investigate the reasons for the users' decisions and their confidence in them. We show that SPISM outperforms other kinds of global and individual policies, by achieving up to 90% of correct decisions.

In this paper, we propose a formal model of coordinated attacks in which several attackers cooperate towards a common malicious goal. The model investigates both attack planning and vulnerability analysis, thereby providing a uniform approach to system and adversary modelling. In addition, the model is general enough to explain both coordinated and single attacks.In the paper, we define the notion of coordinated-attack graph, propose an algorithm for efficient generation of coordinated-attack graphs, demonstrate how coordinated-attack can be used for vulnerability analysis, and discuss an implementation of a coordinatedattack graph.Coordinated-attack graphs can facilitate a wide range of tasks, such as model checking, opponent modelling, intrusion response, sensor configuration, and so forth. In addition, they can be used in robotic warfare, where several intelligent software agents automatically produce and launch coordinated attacks.