Current data collection protocols for wireless sensor networks are mostly based on quasi-static minimum-cost routing trees. We consider an alternative, highly-agile approach called backpressure routing, in which routing and forwarding decisions are made on a per-packet basis. Although there is a considerable theoretical literature on backpressure routing, it has not been implemented on practical systems to date due to concerns about packet looping, the effect of link losses, large packet delays, and scalability. Addressing these concerns, we present the Backpressure Collection Protocol (BCP) for sensor networks, the first ever implementation of dynamic backpressure routing in wireless networks. In particular, we demonstrate for the first time that replacing the traditional FIFO queue service in backpressure routing with LIFO queues reduces the average end-to-end packet delays for delivered packets drastically (75% under high load, 98% under low load). Further, we improve backpressure scalability by introducing a new concept of floating queues into the backpressure framework. Under static network settings, BCP shows a more than 60% improvement in max-min rate over the state of the art Collection Tree Protocol (CTP). We also empirically demonstrate the superior delivery performance of BCP in highly dynamic network settings, including conditions of extreme external interference and highly mobile sinks.

The capabilities of sensor networking devices are increasing at a rapid pace. It is therefore not impractical to assume that future sensing operations will involve real time (inelastic) traffic, such as audio and video surveillance, which have strict bandwidth constraints. This in turn implies that future sensor networks will have to cater for a mix of elastic (having no bandwidth constraint requirements) and inelastic traffic. Current state of the art rate control protocols for wireless sensor networks, are however designed with focus on elastic traffic. In this work, by adapting a recently developed theory of utilityproportional rate control for wired networks to a wireless setting, and combining it with a stochastic optimization framework that results in an elegant queue backpressure-based algorithm, we have designed the first-ever rate control protocol that can efficiently handle a mix of elastic and inelastic traffic in a wireless sensor network. We implement this novel protocol in a real world sensor network stack, the TinyOS-2.x communication stack for IEEE 802.15.4 radios and evaluate the real-world performance of this protocol through comprehensive experiments on 20 and 40-node subnetworks of USC's 94-node Tutornet wireless sensor network testbed.

The state of the art congestion control algorithms for wireless sensor networks respond to coarse-grained feedback regarding available capacity in the network with an additive increase multiplicative decrease mechanism to set source rates. Providing precise feedback is challenging in wireless networks because link capacities vary with traffic on interfering links. We address this challenge by applying a receiver capacity model that associates capacities with nodes instead of links, and use it to develop and implement the first explicit and precise distributed rate-based congestion control protocol for wireless sensor networks --- the wireless rate control protocol (WRCP). Apart from congestion control, WRCP has been designed to achieve lexicographic max-min fairness. Through extensive experimental evaluation on the USC Tutornet wireless sensor network testbed, we show that WRCP offers substantial improvements over the state of the art in flow completion times as well as in end-to-end packet delays.

The receiver capacity model is a simple model to capture flow dynamics in a multi-hop wireless network, by presenting linear constraints to define the feasible rate region of the network, taking into account interference. The model associates with each receiver in the network a notion of constant receiver capacity. Receiver capacity is defined as the maximum possible sum rate of all flows that the receiver can send, receive, and overhear. As has been shown in prior work by the authors, the linear constraints presented by this model make it particularly useful in approximating the true rate region, and designing distributed protocols for multi-hop wireless networks. It is well known that if we use only local constraints to define the rate region, the constraints have to be bounded by some fraction of the interference free link rate, in order to ensure that the rate satisfying these constraints can be feasibly scheduled in any graph. The key challenge in using this model is therefore to estimate the fraction of the link rate that the receiver capacity should be set to, in order to present a feasible rate vector. In this work we answer this question from a theoretical standpoint, and show that as long as the receiver capacity is set to1/3the interference free link rate, all rate vectors that satisfy the constraints of the receiver capacity model can be feasibly scheduled.

Considerable research has been done on detecting and blocking portscan activities that are typically conducted by infected hosts to discover other vulnerable hosts. However, the focus has been on enterprise gateway-level intrusion detection systems where the traffic volume is low and network configuration information is readily available. This paper investigates the effectiveness of existing portscan detection algorithms in the context of a large transit backbone network and proposes a new algorithm that meets the demands of aggregated high speed backbone traffic. Specifically, we evaluate two existing approaches - the portscan detection algorithm in SNORT, and a modified version of the TRW algorithm that is a part of the intrusion detection tool BRO. We then propose a new approach, TAPS, which uses sequential hypothesis testing to detect hosts that exhibit abnormal access patterns in terms of destination hosts and destination ports. We perform a comparative analysis of these three approaches using real backbone packet traces, and find that TAPS exhibits the best performance in terms of catching the maximum number of true scanners and yielding the least number of false positives. We have a working implementation of TAPS on our monitoring platform. Further implementation optimizations using bloom filters are identified and discussed.

When the data rates in sensor networks are comparable to the available channel bandwidth, traditional randomized access schemes face the problem of energy inefficiency and reduced throughput due to increased MAC collisions as well as the problem of unfair data delivery. We argue that under such conditions it is preferable to focus on techniques for scheduled access. We present a linear programming formulation and corresponding distributed TDMA-based scheduling algorithms to provide max-min fair collision-free bandwidth allocation to all sources. We evaluate the performance of the proposed scheduled flow technique using the Tossim/Nido network simulator for the Berkeley Mote/TinyOS platform. Our results show that under high data rate conditions, the proposed scheme significantly outperforms randomized access based schemes in terms of key metrics such as fairness, energy efficiency, throughput, and delay.