Intelligent search in social communities of smartphone users

Andreas Konstantinidis, Demetrios Zeinalipour-Yazti, Panayiotis Andreou, George Samaras, Panos K. Chrysanthis
2012 Distributed and parallel databases  
Social communities of smartphone users have recently gained significant interest due to their wide social penetration. The applications in this domain, however, currently rely on centralized or cloud-like architectures for data sharing and searching tasks, introducing both data-disclosure and performance concerns. In this paper, we present a distributed search architecture for intelligent search of objects in a mobile social community. Our framework, coined SmartOpt, is founded on an in-situ
more » ... a storage model, where captured objects remain local on smartphones and searches then take place over an intelligent multi-objective lookup structure we compute dynamically. Our MO-QRT structure optimizes several conflicting objectives, using a multi-objective evolutionary algorithm that calculates a diverse set of high quality non-dominated solutions in a single run. Then a decision-making subsystem is utilized to tune the retrieval preferences of the query user. We assess our ideas both using trace-driven experiments with mobility and social patterns derived by Microsoft's GeoLife project, DBLP and Pics 'n' Trails but also using our real Android SmartP2P 3 system deployed over our SmartLab 4 testbed of 40+ smartphones. Our study reveals that SmartOpt yields high query recall rates of 95%, with one order of magnitude less time and two orders of magnitude less energy than its competitors. System Model and Problem Formulation In this section, we outline our system model and formulate the problem SmartOpt aims to solve. A table of respective symbols is shown in Table 1 . System Model Overview: Let C, denote a social networking service that maintains centrally the profiles P = {p 1 , p 2 , ..., p M }, for each of its M subscribed users (i.e., U = {u 1 , u 2 , ..., u M }). The profiles record basic user details, authentication credentials, the user interests (e.g., 6 traveling, sports, music, etc.) and friendship relations that define the conceptual social network graph G among the M users. In our setting, a user u i (i ≤ M ) uses a smartphone (or tablet) device to both perform its day-to-day activities but also to capture objects of interest at arbitrary moments (e.g., "take a picture of the Liberty Statue"). Each object o ik might be tentatively "tagged" with GPS information and other user tags (e.g., "lat: 40.689201355, long: -74.0447998047, tags: "Statue Liberty Ellis Island"). Connection Modalities: Each u i features different Internet connection modalities that provide intermittent connectivity to C (e.g., WiFi, 2G/3G/4G). Each u i also features peer-to-peer connection modalities that provide connectivity to nodes in spatial proximity (e.g., Bluetooth, Portable WiFi or upcoming NFC available in Android). We assume that when u i is connected to C, then C is aware of u i 's absolute location (e.g., GPS) or u i 's relative location (e.g., the cell-ids within u i 's range, WiFi RSSI indicators within u i 's range or other means utilized for geo-location). Notice that each of the connection modalities comes at different energy and data transfer rate characteristics. For example, we've profiled an Android-based HTC Hero and found that WiFi consumes 39mW/byte, 3G consumes 24mW/byte and Bluetooth consumes 14mW/byte. Additionally, Bluetooth had a symmetric data rate of 864kbps, while WiFi an asymmetric data rate of 123Kbps (up) and 2Mbps (down) and 3G an asymmetric data rate of 2.7Mbps (up) and 7.2Mbps (down). The nominal data rates for the aforementioned modalities might differ significantly, as this is also validated in [21] , mainly due to the deployment environment. Moreover, while the power consumption on the different kinds of radios can be comparable, the energy usage for transmitting a fixed amount of data can differ an order of magnitude because the achievable data rates on these interfaces differ significantly [35] . Finally, the availability characteristics of these kinds of modalities can vary significantly. The penetration of some form of cellular availability (e.g., WiFi or 3G) is significantly higher than Bluetooth, on average. Thus, uploading or downloading large data items using Bluetooth can be more energy-efficient than using a radio network, but Bluetooth may not always be available and it is often slower. Search Techniques: Let an arbitrary user u j (j ≤ M ), be interested in answering a query 10 Q over its social neighborhood (i.e., nodes connected to u j either directly or through intermediate nodes) G (G ⊆ G). For instance, let Q be a depth-bounded breadth first search query over u j 's neighbors in the G graph (i.e., in G ). This kind of conceptual query can be realized in the following manners:
doi:10.1007/s10619-012-7108-0 fatcat:ok5aqb5janeozh7akgqwosv7eu