Research: Cellular Networks

Back-haul of 3G networks:
Current 3G base-station backhaul still relies on T1/T3 lines, which quickly get overwhelmed due to recent exploding smartphone data usage. In this project, we designed a novel resource provisioning approach for reducing bandwidth consumption at the back-haul of 3G base-stations by utilizing user mobility and data usage patterns. 

Our approach is based on the following insights: (i) Users upload content from a small number of locations, typically corresponding to their home or work locations; (ii) because such locations are different for different users, we find that the problem appears ubiquitous, since user-generated content uploads grow exponentially at most locations. However, we also find that (iii) there exists a significant lag between content generation and uploading times. For example, we find that 55% of content that is uploaded via mobile phones is at least 1 day old. Based on the above insights, we propose a new cellular network architecture. Our approach proposes capacity upgrades at a select number of locations called Drop Zones. Although not particularly popular for uploads originally, Drop Zones seamlessly fall within the natural movement patterns of a large number of users. They are therefore better suited for uploading larger quantities of content in a postponed manner. We design infrastructure placement algorithms and demonstrate that by upgrading only about 17% of base-stations of the  nationwide infrastructure of a studied cellular network provider and assuming users would postpone content delivery by 1 day, the analyzed provider can become capable of absorbing 50% of user generated content delivered in a postponed manner as part of the user daily movement. Our paper on this topic is due to appear at ACM/IEEE INFOCOM 2011 in Shanghai in April 2011.

Localization of Cellular Infrastructure:
There is an existing ecosystem of services such as Skyhook, Navizon, etc., which collect information about the physical location of cellular basestations and Wi-Fi hotspots for providing location-based services. Often these services accomplish the same via “war-driving” the entire region of interest and collecting geo-spatial data via a GPS-enabled smartphone, thereby associating the basestations with the geo-location of the “war-driving” user. Such an approach is both expensive and prone to the frequent churn in infrastructure - coverage expansion, mergers-and-acquisitions of cellular services providers (CSP) etc. In this project, one of our primary goals is to provide an efficient (passive) alternative to such a “war-driving” approach. In this project, we propose and explore a novel approach to map the cellular infrastructure via explicit user geo-intent. By geo-intent, we mean geo-location information specified by users while submitting queries to certain services (e.g. weather or map services), in which they explicitly seek information regarding a specific location. We develop techniques to correlate users' geo-intent with the base-station that they are associated with and show how we can infer the location of a base-station quite accurately by using user queries. Moreover, in some cases, we can even infer the location of even those base-stations that we didn't see a user query to be associated with, solely on the basis of users' movement patterns. More details soon in the paper that will appear at ACM/IEEE INFOCOM 2011 MiniConference in Shanghai in April 2011.

Location Based Services: 
Characterizing the relationship that exists between people’s application interests and mobility properties is the core question relevant for location-based services, in particular those that facilitate serendipitous discovery of people, businesses and objects. In this paper, we apply rule mining and spectral clustering to study this relationship for a population of over 280,000 users of a 3G mobile network in a large metropolitan area. Our analysis reveals that (i) People’s movement patterns are correlated with the applications they access, e.g., stationary users and those who move more often and visit more locations tend to access different applications. (ii) Location affects the applications accessed by users, i.e., at certain locations, users are more likely to evince interest in a particular class of applications than others irrespective of the time of day. (iii) Finally, the number of serendipitous meetings between users of similar cyber interest is larger in regions with higher density of hotspots. Our analysis demonstrates how cellular network providers and location based services can benefit from knowledge of the inter-play between users and their locations and interests. For more details please refer to our paper titled “Measuring Serendipity: Connecting People, Locations and Interests in a Mobile 3G network" that appeared in ACM Internet Measurement Conference (IMC) 2009, Chicago.

3G worms:
Recently, cellular phone networks have begun allowing third-party applications to run over certain open-API phone operating systems such as Windows Mobile, iPhone and Google’s Android platform. However, with this increased openness, the fear of rogue programs written to propagate from one phone to another becomes ever more real. This paper proposes a counter-mechanism to contain the propagation of a mobile worm at the earliest stage by patching an optimal set of selected phones. The counter-mechanism continually extracts a social relationship graph between mobile phones via an analysis of the network traffic. As people are more likely to open and download
content that they receive from friends, this social relationship graph is representative of the most likely propagation path of a mobile worm. The counter mechanism partitions the social relationship graph via two different algorithms, balanced and clustered partitioning and selects an optimal set of phones to be patched first as those which have the capability to infect the most other phones. The performance of these partitioning algorithms is compared against a benchmark random partitioning scheme. Through extensive trace-driven experiments using real IP packet traces from one of the largest cellular networks in the US, we demonstrate the efficacy of our proposed counter-mechanism in containing a mobile worm. More details can be found in the related paper titled “A Social Network Based Patching Scheme for Worm Containment in Cellular Networks” that appeared at ACM/IEEE INFOCOM 2009, Rio de Janeiro.
Comments