Presentation is loading. Please wait.

Presentation is loading. Please wait.

Shravan Gaonkar, Jack Li, Romit Roy Choudhury, Landon Cox, Al Schmidt Presented by : Gregory Teodoro.

Similar presentations


Presentation on theme: "Shravan Gaonkar, Jack Li, Romit Roy Choudhury, Landon Cox, Al Schmidt Presented by : Gregory Teodoro."— Presentation transcript:

1 Shravan Gaonkar, Jack Li, Romit Roy Choudhury, Landon Cox, Al Schmidt Presented by : Gregory Teodoro

2 A brief overview  Mobile Computing  Hardware is increasing in power, and decreasing in cost such that basic commodity devices will be able to perform work such as participatory sensing.  Mobile phones especially, can be used by leveraging their camera, GPS, accelerometers, and other such sensors to gather information.  Using a multitude of these phones over a large area it is possible to build a “virtual information telescope”. A visual large scale interface, that can quickly shift to small and more detailed specifics.  Combined with these sensors, participating users can record multimedia blogs (video, voice, pictures, ect.).  Deeper in, these blogs can be combined with a user-based query/answer. (ala services such as Cha-Cha, but less personal and more public/social)  Combined, all of these factors form…

3 …the Micro-Blog Service!  Micro-blog stands to combine all of this information  Idea being to create a network of easily accessible, localized information.

4 Services - Querying  Querying Service  The Micro-blogs main goal is to offer a querying services, through use of the influx of sensor information, a user can gather information about temperature, weather, available Wi-Fi services  Queries can also require human interaction.  Questions can be sent out to an area about nearby restaurants, locations, the weather, attractions ect… and then answered by people in the area.

5 Services – Location-Based  Queries and blogs can also be superimposed on a virtual space (map) or a physical space (a part of a city). “Float”.  Person A sends out micro-blog about Restaurant X in Area Y.  Person B comes to Area Y and has set an interest in information on “Restaurants”. Since he’s in Area Y, he’d receive Person A’s microblog about the Restaurant X.  Relies on social participation.

6 Potential Applications  Tourism  On the fly information about attractions in an area, either through blogs or audio/videos.  Can ask questions about services in an area (parking, hotels, ect)  News Services  Users send out and report events they see, or sensors information can create and offer weather announcements.  Spontaneity of capturing information and easier publishing means immediate news reports

7 Potential Applications (cont.)  Micro-Alerts  Location-aware alerts, an example being a college where microblogs can be placed directing users to specific buildings.  Comparable best to floating sticky notes.  Social Collaboration  Ability to offer on the fly services, ala Craigslist.  Car Pools and Healthcare Reports for example.  Can post queries and questions, and qualified users can answer them.

8 Architecture and Design  Location based Micro-blogs require a localization service.  GPS, GSM, WiFi, ect  Issue of potential power consumption and accuracy balance.  Blog are location-tagged then indexed by originator, access permissions, themes, ect.  Client contacts webserver to access and retrieve a micro-blog.  Users can also send queries to specific regions R.  Using the localization, the database can see if it has any serviceable blogs to send to the user, and attempts to match time, location, and permissions.  If it cannot find an answer, it will send a query to phones in the area that are declared available, and they may answer the query.  Any answers are placed onto the map as a new micro-blog, and may be deleted after a pre-specified lifetime.  Users will also automatically received micro-blogs pertaining to their interests and the region they are in, triggered when the phone updates its location.

9 Block-Diagram of Micro-Blog

10 Considerations  Energy Aware Localization  Constant running of GPS localization is a heavy drain on energy, and may not be required for acceptable accuracy.  Energy Aware Applications  Applications should adapt to energy needs and efficiency versus performance.  Incentives  Why would people answer queries if they don’t have to?  Location Privacy  Location-based blogging reveals the bloggers general location, security and privacy concerns.  Spam  Malicious users may use the service to spam other users.  Content Inaccuracy  Not all query answers might be correct, and malicious users may outright lie.

11 Energy Awareness  Rather than using a single localization scheme, we can use a mix.  Use GPS to find the general location once, then use the less expensive WiFi or GSM localizations to keep track.  We only use GPS if we have too, or if the user has traveled a distance above a certain threshold.  Using a technique called war-driving, a map of WiFi Aps can be created, and then used (by detecting if a user is in range) to device a users approximate location.

12 Interpolated Path

13 Problems with this…  Does not achieve optimal usage.  Sharp turns in a path may not trigger a change in the WiFi-estimated distances.  Will continue to extrapolate the original direction of movement, and not the sudden turns.  Will not correct itself until a new WiFi fingerprint is created.

14 Energy Aware Applications  Only updates by GPS when required, or when a blog is actually created and placed.  When not created a blog, it will only updated based on Wi-Fi or cell-based localizations, as exact location or accuracy may not be required.  Delay updating and uploading blogs if battery power is below a certain threshold.  Keeps power on the phone in case of an unforeseen emergency.

15 Incentives  Why should people answer queries if they don’t need to?  Queries can be restricted to social network  A user’s friends would be far more likely to answer than a stranger.  Alternatively a give and take approach  Users would be given “query credits” based on how often they participate and answer other user’s queries.  Issues with bogus queries, collusion, cliques, ect.  Malicious or misbehaving users can be found through use of a direction node with a cost representation how often the two nodes query each other.  Monetary costs for sending out queries and answers  Plan is to piggy-back on a value package offered by phone providers.

16 Locational Privacy  Location-based blogging reveals the bloggers location!  A trust is required between the users and the server administrators at Micro-Blog. A fundamental drawback.  They make the assumption that users are willing to trust Micro-Blog.  Offers a private, social, and a public mode as an alternative.

17 Spam  The network can be used send spam to users  Proposed solution is to limit the amount of queries a single user can receive from each group (private, social, public) during a time period.  Blocks any queries past this point  Does not solve the issue!  A user can be spammed to use up the number of queries they want to receive a day, preventing any future legitimate queries from being received.

18 Content Inaccuracy  Not all information will be correct.  Users will receive a reputation based on other user’s ratings.  Akin to other social services like Digg.  The higher a reputation, the more confidence the micro- blog will have that their query answer is correct.  Do not publish answers from sources that are not confident.  PROBLEMS!  A malicious user will be quickly locked out, as will unknowledgeable, but they will still be able to send out some disinformation  No methods to stop a malicious user from acting legitimate to gather a good reputation, and then abusing it  Ties into group/user collusion as well.

19 Implementation  The Micro-Blog program is placed onto a phone client.  Designed specifically to interrupt the normal operations of a phone as little as possible.  System State  System begins in Idle mode  User goes to send a blog  When finished, blog is sent out, phone is tagged with location, and then switches to Sync mode  Sync Mode  Phone syncs information with the webserver and receives any new, relevant Micro-blogs, then returns to Idle  Log State  Entered periodically where phone updates its status based on sensor data (signal strength, temperatures, cameras, WiFi/GSM IDs, ect), then returns to Idle  Query state  A query was receive and appears on the users phone and waits for an action. Either reply or ignore. Sends it out, then returns to Idle  This is also checked periodically.

20 State Machine for Micro-Blog

21 Web Infastructure  Phone data is send to a database via TCP, and uploading into a MySQL database server, then indexed into proper relational tables.  Blogs in response to an earlier query are sent out to the approriate users.  Apache 2.0 is used to offer HTTP access to the Micro-Blog over a web browser, where users can request blogs, create blogs, and browse a region map.  Server periodically checks all phones that have a live connection and their user-set constraints and send queries to them as appropriate.

22 Battery Life and Localization  A mix between localization techniques, and the use of battery must be found.  Findings were done using a Nokia N95 phone.  Main research was done through the use of localization trades off.  Localization schemes included using combinations of WiFi, GPS, and GSM.

23 Findings

24 Findings (cont.)  Battery life improved using a mix of the localization schemes as appropriate.  This still had a negative effect on locational accuracy  As far as the application is concerned the trade-off of accuracy is worth the improvement of battery life  Methods Attempted  Wifi with periodic GPS correction  GSM with periodic WiFi Corrections  Proposed accuracy error was a max of 125m, leading to a total of 25 hours of battery life.  GSM with periodic WiFi offered the best battery life vs. localization error, and is used in the Micro-Blog program.

25 Findings (Cont.)

26 Reception  Questionnaires were done on participants about the application and how they felt about it.  Great for idling periods, but needs a better GUI.  Privacy control is absolutely vital, and users appreciated the ability to control privacy.  Incentives may not be an issue if program is “fun”.  Strong correlation between voice blogs being personal, and text blogs being impersonal.  Best time of day is between 5:00-9:00 PM  Concerns about battery life minimal because recharging is easy or imminent.

27 Paper Strengths  Very good ideas about what the software should do, and some basic logistics behind it.  Highly concerned with battery and location accuracy, and answers the question about what should be used well.  Lots of future potential in the subject at hands.  Does attempt hands-on testing and user-base interaction.

28 Paper Weaknesses  Ignores a lot of potential networking issues, such as sorting through data, dealing with traffic, packet loss, congestion, among other things  Glazes over some of the privacy and malicious user issues, and does not fully solve them.  Though it was a user study, not many users were involved (only 12), nor were these users varied in location, culture, or age.  Needs a far more extensive study of a user-base  Paper does admit this as a fault of the study.

29 Future Work/Follow Ups  Location-Privacy in the General Public  A far broader audience needs to be approached with the software and tested.  Accelerometer usage  Accelerometers are not used in terms of gauging distance travel and localization techniques  Has potential to be integrated into social networking applications (Facebook, MySpace, Google+)  Methods of identifying and blocking false content need to be developed.

30 Questions?


Download ppt "Shravan Gaonkar, Jack Li, Romit Roy Choudhury, Landon Cox, Al Schmidt Presented by : Gregory Teodoro."

Similar presentations


Ads by Google