Cedric Angelo M. Festin

A comparison of Wireless Fidelity (Wi-Fi) fingerprinting techniques

Among several techniques proposed for indoor positioning using IEEE 802.11 Wireless Fidelity (Wi-Fi) based networks, those that rely on fingerprinting have been demonstrated to outperform those based on lateration, angulation, and cell of origin in terms of accuracy. We compare and evaluate three Wi-Fi fingerprinting techniques that use the K-Nearest Neighbor (k-NN), Naive Bayes Classifier (NBC) and Support Vector Machine (SVM) algorithms. Our experiments show that SVM-based fingerprinting outperformed both k-NN and NBC-based fingerprinting, achieving accuracies of 2 meters or better within our testbed.

Utility-based buffer management for networks

User satisfaction from a given network service or resource allocation can be viewed as having two aspects, a state and a degree. The state defines whether the user is happy or unhappy. A user is happy when its expectations are met. The degree defines the level of happiness or unhappiness. We present the use of perceived knowledge of the state and degree of user satisfaction in managing router resources and functions and examine whether such knowledge could help a router improve local resource allocation decisions. We describe our formulation, Value-Based Utility (VBU) that incorporates both aspects of user satisfaction. We establish a framework of VBU use and demonstrate its application to buffer management. We propose a FIFO scheme that uses VBU and evaluate its success in meeting user expectations. The main conclusion we draw from this work is that the VBU framework offers a different perspective of performance definition and analysis and allows for the effective distribution of resources especially in times of high demand and low resource availability. Its adoption into existing traffic management schemes is further motivated by the improved performance of our proposed scheme over its non-VBU aware counterpart.

An Investigation of Phone Upgrades in Remote Community Cellular Networks

In the last decade, billions of people worldwide have upgraded from basic 2G feature phones to data-enabled 4G smartphones. In most cases, people upgrade in areas with 4G coverage (typically cities and large towns), but increasingly, people choose to upgrade in areas that only have 2G coverage or no cellular coverage at all. This counterintuitive behavior -- upgrading your phone despite living in an area that does not actively support many of the features of that new device -- is the focus of this work. We investigate the rates and reasons for 4G upgrades and adoption in two extremely remote areas in Indonesia and the Philippines. Our mixed-methods approach combines the quantitative analysis of several years of mobile phone registration logs with the qualitative analysis of multiple interviews in one of these communities. We learn that users are rapidly switching from 2G to 4G technology and skipping 3G entirely; the data suggest that these villages will soon have sufficient 4G phone adoption to justify the investment required to upgrade base stations to 4G technology. The interviews suggest people are making these switches primarily to support consumption of media such as games, videos, and music. Similarly, users switch devices because of damage, often leading to downgrades to more resilient feature phones. We also find that, despite the general value seen in more modern 4G phones, 2G phones are more shared and more active on the network.

Multipath Bandwidth Scavenging in the Internet of Things

In order to achieve the capacity and geographic scope required by IoT applications, service providers should explore mutually-beneficial modes of collaboration such as through cooperative packet forwarding by IoT nodes and cooperative gatewaying through fixed backhauls. To promote such resource pooling while minimizing negative impact on collaborating providers, we developed a transport-layer approach that combines multipath techniques with less-than-best effort (LBE) congestion control methods to enable IoT nodes to opportunistically scavenge for idle bandwidth across multiple paths. Initial tests using TCP-LP and LEDBAT congestion control algorithms on scavenging secondary flows show that this desired functionality can be achieved, while our use of standard TCP congestion control on primary flows ensures that IoT nodes are guaranteed at least one flow that can compete for fair share of the network capacity.

A Hybrid Incentive-based Peer Selection Scheme for Unstructured Peer-to-Peer Media Streaming Systems

The success of peer to peer file sharing protocols such as BitTorrent and Gnutella make peer to peer networks an attractive alternative for implementing media streaming services. However, due to the continuous nature of the content, the key components of such conventional file sharing applications are unable to suffice to the more stringent requirements of media streaming. In this study, we propose an alternative to one of its key components that is the incentive based peer selection routine. The primary motivation of our proposed scheme is the utilization of peer contribution information on both local and global context for a more informed peer selection process.

Evaluating the effects of peer localization on a BitTorrent-based P2P video-on-demand network

Video-on-demand (VoD) is a rapidly growing method for viewing and sharing video content over the Internet. However, video on demand consumes a significant amount of bandwidth, forcing content publishers and content delivery networks to examine approaches such as the use of peer-to-peer (P2P) technologies to assist in video distribution. We study the performance impact of a peer localization strategy aimed at limiting traffic exchanges between autonomous systems such as Internet service providers (ISPs) in a P2P VoD setup. Our findings suggest that peer localization may improve streaming performance while reducing network traffic crossing autonomous systems, potentially benefitting both users and providers, respectively.

Towards Building a Community Cellular Network in the Philippines: Initial Site Survey Observations

In this paper, we present preliminary site survey findings for a community cellular network project in the Philippines. We document our observations from our visits to unserved communities, notably on how they deal and adapt to the lack of cellular access. We also identify challenges in deploying the technology in the Philippine setting.

Tangible Sharing, Invisible Mechanisms: The Design and Implementation of the BayanihaNets Access Sharing Network

Cooperation through resource pooling may help deliver affordable connectivity to those without subscriptions and improve access bandwidth for those with existing connections. Using the concept of *bayanihan*, a Filipino expression for a deeply traditional and cultural type of cooperation, we designed and implemented the *BayanihaNets* access sharing network, a cooperative system that delivers connectivity through the practice of *bayanihan* among its users. It induces *bayanihan* by making sharing tangible to the users, through mechanisms that highlight the altruism they exercise in sharing connectivity and the negative impact of selfish actions, if any. Our strategy of promoting and enabling network resource sharing as a social activity is achieved through the use of the familiar interfaces, mechanisms and idiom from social networking applications: our approach may thus be described as social networking through social networking. We discuss our *bayanihan*-based design principles and present our initial experiments on bandwidth sharing in a controlled environment.

Peering into peering: Building better tools for better peering decisions

Network operators need to assess the effects of routing policies and traffic engineering methods in order to guide planning and operational decisions related to peering. We propose a peering analysis framework based on the correlation of inferred AS paths and traffic flow information. We define the data sets needed and a four-step methodology designed to reduce network data dimensionality, determine traffic proximity, identify traffic propensity, and quantify the impact of traffic locality. We demonstrate how the correlation of traffic flow and AS paths, and the application of our four-step approach, uncovers rich information when applied to a real-world case study of a local Internet service provider.

Delay-based end-to-end congestion control for wireless sensor networks

In wireless sensor networks, commonly used networking protocols may not necessarily function well due to the stringent requirements of such networks. A transport layer protocol for this environment must work well over shared wireless connections that are prone to packet errors. TCPs burstiness due to its aggressiveness in probing for available capacity and multiplicative backoff in response to congestion yields a traffic profile that may not be suitable for wireless sensor networks. This work investigates how LEDBAT, a delay-based less-than-best-effort protocol, would behave in a wireless sensor network. Although LEDBAT was designed to yield to TCP when sharing a bottleneck link, simulations showed that under certain conditions, LEDBAT obtains a fair share of wireless bandwidth, ensuring node throughput in WSN scenarios. More importantly, it is also able to provide a steadier flow than TCP even as the number of hops to its destination increases.