When filling out privacy-related forms in public places such as hospitals or clinics, people usually are not aware that the sound of their handwriting leaks personal information. In this paper, we explore the possibility of eavesdropping on handwriting via nearby mobile devices based on audio signal processing and machine learning. By presenting a proof-of-concept system, WritingHacker, we show the usage of mobile devices to collect the sound of victims' handwriting, and to extract handwriting-specific features for machine learning based analysis. WritingHacker focuses on the situation where the victim's handwriting follows certain print style. An attacker can keep a mobile device, such as a common smartphone, touching the desk used by the victim to record the audio signals of handwriting. Then, the system can provide a word-level estimate for the content of the handwriting. To reduce the impacts of various writing habits and writing locations, the system utilizes the methods of letter clustering, dictionary filtering and letter time length based offsetting. Moreover, if the relative position between the device and the handwriting is known, a hand motion tracking method can be further applied to enhance the system's performance. Our prototype system's experimental results show that the accuracy of word recognition reaches around 70 - 80 percent under certain conditions, which reveals the danger of privacy leakage through the sound of handwriting.

Indoor localization based on Wi-Fi fingerprints has been an active research topic for years. However, existing approaches do not consider the instability of access points (APs) which may be unreliable in practice, particularly the ones deployed by individual users. This instability impacts the localization accuracy severely, due to the unreliable or even wrong Wi-Fi fingerprints. Ideally, the localization should be done using only the well-deployed APs (e.g., deployed by facility teams). However, in many places the number of these APs is too few to achieve a good localization accuracy. To solve this problem, we leverage emerging smart APs equipped with multi-mode antennas, and build a new indoor localization system called MMLOC to reduce the number of necessary APs. The key idea is controlling the modes of AP antennas to generate more fingerprints with fewer APs. A clustering based localization strategy is designed to enable a mobile terminal to figure out the RSSI (Received Signal Strength Indicator) for different antenna modes without requiring any synchronization. We have implemented a prototype system using smart APs and commercial smartphones. Experimental results demonstrate that MMLOC can reduce the number of necessary APs by 50%, and achieve the same or even better localization accuracy.

Currently, the surveillance camera-based person re-identification is still challenging because of diverse factors such as people’s changing poses and various illumination. The various poses make it hard to conduct feature matching across images, and the illumination changes make color-based features unreliable. In this article, we present SKEPRID, 1 a skeleton-based person re-identification method that handles strong pose and illumination changes jointly. To reduce the impacts of pose changes on re-identification, we estimate the joints’ positions of a person based on the deep learning technique and thus make it possible to extract features on specific body parts with high accuracy. Based on the skeleton information, we design a set of local color comparison-based cloth-type features, which are resistant to various lighting conditions. Moreover, to better evaluate SKEPRID, we build the PO8LI 2 dataset, which has large pose and illumination diversity. Our experimental results show that SKEPRID outperforms state-of-the-art approaches in the case of strong pose and illumination variation.

Indoor localization for quadrotors has attracted much attention recently. While efforts have been made to perform location estimation of quadrotors leveraging dedicated indoor infrastructures, the low-cost and commonly used RSS fingerprinting based approach utilizing existing Wi-Fi APs has yet to be applied. The challenge is that the high-speed flight reduces the RSS measuring opportunities for fingerprints comparison; moreover, the 3D space fingerprints collection incurs more overhead than in the traditional 2D case. In this paper, we present HiQuadLoc, a RSS fingerprinting based indoor localization system for quadrotors. We propose a series of mechanisms including path estimation, path fitting, and location prediction to deal with the negative influence incurred by the high-speed flight; moreover, we develop a 4D RSS interpolation algorithm to reduce the site survey overhead, where 3D is for the indoor physical space and 1D is for the RSS sample space. Experimental results demonstrate that HiQuadLoc reduces the average location error by more than 50 percent compared with simply applying the RSS fingerprinting based approach for 2D localization, and the overhead of RSS training data collection is reduced by more than 80 percent.

When filling out privacy-related forms in public places such as hospitals or clinics, people usually are not aware that the sound of their handwriting leaks personal information. In this paper, we explore the possibility of eavesdropping on handwriting via nearby mobile devices based on audio signal processing and machine learning. By presenting a proof-of-concept system, WritingHacker, we show the usage of mobile devices to collect the sound of victims' handwriting, and to extract handwriting-specific features for machine learning based analysis. WritingHacker focuses on the situation where the victim's handwriting follows certain print style. An attacker can keep a mobile device, such as a common smart-phone, touching the desk used by the victim to record the audio signals of handwriting. Then the system can provide a word-level estimate for the content of the handwriting. To reduce the impacts of various writing habits and writing locations, the system utilizes the methods of letter clustering and dictionary filtering. Our prototype system's experimental results show that the accuracy of word recognition reaches around 50% - 60% under certain conditions, which reveals the danger of privacy leakage through the sound of handwriting.

The contradiction between dynamic user traffic and fixed data plans has drawn increasing attention in the field of mobile applications. In this paper we build a data plan sharing system named INDAPSON to consider a scenario where some smartphone users have surplus data traffic and are willing to help others download data. Virtual credits can be gained as reward, which can be used to ask for future help in downloading. To realize this model, we make the following contributions: 1) A dynamic self-organization strategy to adapt to mobile terminals; 2) An incentive mechanism named RAP to encourage participation; 3) Power-saving strategies to reduce power consumption. The main advantage of our system is that users gain improvement in download rate while being able to convert their surplus data traffic to virtual credits. The results of experiment and simulation show that users in our system can manage their surplus data plan more efficiently while a highspeed download rate can be achieved.

There have been many traffic light control systems around the globe, but the high cost of infrastructure and maintenance hinders their wide deployment. However, speed-advisory systems enabled by on-vehicle devices are much cheaper and easier to deploy. The first challenge of such systems is to get the traffic signal schedule in complex intersections. The second challenge is to get map information and calculate the distance. Facing these challenges we devise and implement a speed-advisory driving system called CityDrive, which harnesses the sensor and GPS data from a wide participation of smartphones to suggest proper speed for drivers so that they arrive at intersections in green phase. CityDrive first generates a road map and then infers traffic signal schedules, using only smartphones and a server. CityDrive does not eliminate stops at intersections, but it tries to maximize the probability that vehicles cruise through intersections in green phase. Both simulation and real test show that this continuous speed advisory service effectively smoothes traffic flow and significantly reduces energy consumption.

We study the fundamental lower bound for node buffer size in intermittently connected wireless networks. The intermittent connectivity is caused by the possibility of node inactivity due to some external constraints. We find even with infinite channel capacity and node processing speed, buffer occupation in each node does not approach zero in a static random network where each node keeps a constant message generation rate. Given the condition that each node has the same probability p of being inactive during each time slot, there exists a critical value pc(λ) for this probability from a percolation-based perspective. When pc(λ), the network is in the supercritical case, and there is an achievable lower bound (In our paper, “achievable” means that node buffer size in networks can achieve the same order as the lower bound by applying some transmission scheme) for the occupied buffer size of each node, which is asymptotically independent of the size of the network. If p > pc(λ), the network is in the subcritical case, and there is a tight lower bound Θ(√n) for buffer occupation, where n is the number of nodes in the network.