Data Dissemination Protocols in Wireless Sensor Networks : Models, Security and Design

1
Data Dissemination Protocols in Wireless Sensor
Networks Models, Security and Design

• Candidacy Proposal Defense by Pradip De
• Department of Computer Science and Engineering
• The University of Texas, Arlington
• Center for Research in Wireless Mobility and
  Networking

Advisors Sajal K. Das and Yonghe Liu Committee
Members Kalyan Basu, Mohan Kumar, Matthew Wright
2
Organization

• Data Dissemination in Wireless Sensor Networks
• Protocol Models and Performance Analysis
• Security Analysis of Data Dissemination in Sensor
  Networks
• Design of a Reprogramming Protocol for Mobile
  Sensor Networks
• Ongoing and Future Work

3
Data Dissemination/Reprogramming in Wireless
Sensor Networks
4
Motivation

• Sensor Networks operate unattended for months or
  years
• Little control once deployed
• Environment changes over time
• Reprogramming of sensors is essential
• Evolving requirements
• Bug fixes after deployment
• Scale and embedded nature requires network code
  propagation
• Data dissemination protocols

5
Protocol Design Objectives and Issues

• Scalability
• Size of network
• Data/Code Size disseminated
• Reliability
• Robust against packet loss (wireless
  uncertainties)
• Tolerate topology changes (node failures)
• Reach all nodes staying uncorrupted
• Efficiency
• Rapid propagation required
• Security
• Authentication necessary

6
Protocols in the Horizon Trickle Levis et al,
NSDI 2004 Deluge Hui and Culler, SenSys
2004 MNP Kulkarni and Wang, ICDCS 2005
7
Deluge
J. W. Hui and D. Culler. The dynamic behavior of
a data dissemination protocol for network
programming at scale. In Proceedings of the
second International Conference on Embedded
Networked Sensor Systems (SenSys 2004)
8
Deluge Protocol Overview

• A General Protocol for Bulk Data Dissemination
• State Machine with strictly local rules
• Nodes advertise, request data, and broadcast
• Object divided into contiguous pages, each
  consisting of N packets
• Allows for spatial multiplexing

9
Objectives of Dissertation

• Model based performance analysis of data
  propagation over dissemination protocols
• Rate of information propagation

• Model based security analysis of network-wide
  dissemination
• Spread of node compromise in a sensor network
  with secure communication using pairwise keys
• Malware propagation over data dissemination
  protocols

• Design of a reprogramming protocol for mobile
  sensor networks
• Performance analysis of protocol in mobile
  scenario
• Modeling data/malware propagation over
  dissemination protocols in mobile sensor networks

10
Model Based Comparative Performance Analysis of
Data Dissemination Protocols
11
Model Characteristics and Features

• An epidemic theoretic model for analysis of data
  propagation over these protocols
• Analytical tool for studying dissemination
  protocols
• Measures rate of information propagation
• Flexibility of model
• accommodates different dissemination protocols
• Mechanism for inter-protocol comparison
• Propagation speed
• Extent of coverage

12
Sensor Network Model

• Modeled as an undirected geometric random graph
• N nodes uniformly randomly distributed
• Unit Disk Model with transmission radius
• is the probability of edge existence
  between nodes u and v
• at distance
• Node Density
• where A is the area of the terrain

u
v
13
Epidemic Theory Overview

• Epidemic Theory
• Models an infection spread in a population of
  susceptibles
• Broadly two kinds of modeling techniques
• Random Graph based spatial model
• Differential Equation based temporal model
• Infection Spread Cases
• Susceptible-Infected-Susceptible (SIS)
• Susceptible-Infected-Recovered (SIR)
• Homogeneously mixed population
• Heterogeneously mixed population

14
Epidemic Theoretic Framework

• Proposed Framework
• Design the spread model using network
  characteristics
• Adopt differential equation based approach
• Data propagation conforms to No Recovery model
• Local interactions based on transmission range
• Estimate the rate of infection (ß) based on
• Rate of communication paradigm of the broadcast
  protocol
• Infectivity potential (?) of the data

15
Infection Spread Model
Source Node
Susceptible S(t)
Inoperative R(t)
Infective I(t)
16
Model Derivation

• No Recovery Based Infection Model
• Infected nodes cannot be recovered and the
  infection ultimately reaches the whole network
• Formulation of differential equations for I(t)
  and S(t) based on network parameters
• At , I(t) N

where
17
Fitting Broadcast Protocols

• Deluge
• In the maintenance algorithm, the probability
  of node i broadcasting metadata in each time
  interval
• is given by
• where k denotes the advertisement
  threshold in the period
• and denotes the expected
  number of neighbors of a node i
• The expected time for a node to receive metadata
  is calculated using
• The expected time to transmit a page
  in a neighborhood is derived from
  and the infection rate and is given by

where is the infectivity potential of the data
18
Deluge Data Propagation Rate
Simulation
Analytical
19
Summary

• Performance analysis of broadcast protocols
• Speed of propagation of data
• Reachability into network
• Construction of an epidemic model for data
  propagation
• Flexible tool to compare different broadcast
  protocols

20
Security Analysis of Network-Wide Data
Dissemination in Sensor Networks

• Model based security analysis of network-wide
  dissemination
• Spread of node compromise in a sensor network
  with secure communication using pairwise keys
• Malware propagation over data dissemination
  protocols

21
Propagation of Node Compromise in Sensor Networks

• Construction of a model and analysis of the
  spread of node compromise on a sensor network
  based on Epidemic Theory
• Identify point of outbreak of the process in the
  network
• Observe the impact of infectivity duration of a
  compromised node on the process
• Identify critical values of relevant parameters
  to prevent outbreaks

22
Network Model

• Consider two deployment strategies
• a basic uniform random deployment strategy
• A realistic group based deployment strategy
• Adopt the same model for the physical network
• An overlay with key sharing probability q based
  on random pairwise key predistribution

23
Topology Model Group Based Deployment

• A set of 2-dimensional Gaussian Distribution of
  resident points about the deployment point
• g(x,yj) represents the probability of a node
  belonging to group j to reside within
  transmission range of point x,y

24
Topology Model Group Based Deployment

• The probability that a node at (x,y) has l
  neighbors is expressed as Nb(l,x,y)
• Nb(l,x,y) is a function of g(x,yj) and the
  gaussian distributed node location pdf of (x,y)
• The degree distribution p(k) of a node is given
  by
• where

25
Analysis Overview

• Two scenarios
• No recovery once compromised
• Nodes recover
• When nodes do not recover transmissibility is
  expressed only in terms of the infection
  probability
• Essence of node recovery is captured by
  expressing the transmissibility as a function of
  the average duration of infectivity

26
Primary Analysis Results

• Average Cluster size as the epidemic attains
  outbreak proportions
• Average Epidemic size after outbreak results
• Results observed under both scenarios of without
  node recovery and with node recovery

27
Epidemic size with infection probability
28
Summary

• Study of spread of node compromise in sensor
  networks
• Uniform random network model
• Group deployment based network model
• The outbreak points for network-wide compromise
  propagation are affected by the deployment
  strategy

29
Vulnerability of Broadcast Protocols to Malware
Propagation

• Model based security analysis of network-wide
  dissemination
• Spread of node compromise in a sensor network
  with secure communication using pairwise keys
• Malware propagation over data dissemination
  protocols

30
Vulnerability of Broadcast Protocols

• Construct model to estimate vulnerability to
  piggybacked malware spread
• Compromise propagation after a single or few
  nodes compromised by adversary
• No Recovery case
• Use the same model for data propagation
• Infection ultimately spreads to the entire
  network

31
Attack Model
Malware spreads, piggybacked on the broadcast
protocol, passing security verification at each
stage since source was compromised
Broadcast Protocol wavefront pass
authentication Deploy Malicious Code
Compromised Src Authentication keys captured
32
Model Analysis

• Imposition of a simultaneous recovery process
• Parameterized by mean recovery rate of each
  node
• The infection rate is computed from the
  communication rate of the protocol
• Construct differential equations to compute the
  sub-populations I(t), S(t), and R(t)

33
Deluge Spreading Time Comparison
Simulation
Analytical
With Recovery
34
Summary

• Reprogramming protocols are essential for sensor
  networks
• However, they could be carriers for rapid spread
  of malicious code in sensor network
• Analytical tool proposed
• Gain valuable insights into the propagation
  characteristics of malware over different
  broadcast protocols
• Tool is flexible for comparative studies of
  different broadcast protocols

35
Design of Reprogramming Protocols for Mobile
Sensor Networks

• Design of a reprogramming protocol for mobile
  sensor networks
• Performance analysis of protocol in mobile
  scenario
• Modeling data/malware propagation over
  dissemination protocols in mobile sensor networks

36
Reprogramming Protocols for Mobile Sensor Networks

• Numerous applications for mobile sensor networks
• Drawbacks of the existing reprogramming protocols
  for mobile scenarios
• Location uncertainty due to mobility
• Inefficiency of page ordered download
• Dynamic changes in neighborhood node density
• Protocol should take advantage of mobility

37
ReMo

• Salient features
• Based on a periodic metadata broadcast paradigm
• The probability of broadcast is dynamically
  adjusted based on neighborhood density
• Regardless of order, pages are downloaded based
  on availability
• Snoop on neighborhood to construct link quality
  metrics
• Choose neighbors appropriately for requesting
  downloads based on not only best link quality but
  also high potential of code availability

38
Link Characterization
39
Metadata Broadcast

• and are the counts of the metadata
  advertisements that are different and same as
  current node
• The periodic metadata broadcast probability for
  each time slot t is adjusted based on the above
  counts
• Proportional increase in probability on hearing
  different metadata
• Probability decreased aggressively on hearing
  same metadata

40
Protocol Components and Operation

• Page Download Potential (PDP)
• Based on the pages a node can potentially
  download from a neighbor
• Neighbor Link Profile (NLP)
• Aware of the current link quality with each
  neighbor
• Link Quality estimate is updated as a window mean
  exponentially weighted moving average
• Node i selects a neighbor j to send a download
  request based on NLP and PDP of j

41
Comparison of Code Update Completion Time
42
Number of Message Transmissions
43
Number of Message Transmissions
44
Number of Message Transmissions
45
Ongoing and Future Work

• Design of a reprogramming protocol for mobile
  sensor networks
• Performance analysis of protocol in mobile
  scenario
• Modeling data/malware propagation over
  dissemination protocols in mobile sensor networks

46
Performance Analysis of Data Dissemination
Protocols in Mobile Sensor Networks

• Markov Chain based model of the protocol
  operation
• Borrow ideas from MAC protocol analysis
• 802.11 MAC models for backoff schemes
• Derive throughput of data delivery over these
  protocols

47
Modeling of Information Propagation in Mobile
Sensor Networks

• Analytical model for the data propagation rate in
  mobile sensor network
• Vulnerability assessment in a mobile scenario
• Model approach based on epidemic theory
• Assumption of homogeneous mixing among nodes
  possible

48
Implementation of ReMo

• Implementation of ReMo on a testbed of SunSPOTs
• Test the efficacy of the design of ReMo for code
  download under different real world mobility
  conditions
• Implementation in the Java ME Framework
• Compilation, Deployment and Execution using Ant
  scripts

49
Thank You
50
Relevant Publications

• P. De, Y. Liu, and S. K. Das, Modeling Node
  Compromise Spread in Wireless Sensor Networks
  Using Epidemic Theory. In IEEE International
  Symposium on a World of Wireless, Mobile and
  Multimedia Networks (WoWMoM) 2006.
• P. De, Y. Liu, and S. K. Das, Deployment Aware
  Modeling of Node Compromise Spread in Wireless
  Sensor Networks , under review in ACM
  Transactions on Sensor Networks.
• P. De, Y. Liu, and S. K. Das, Evaluating
  Broadcast Protocols in Sensor Networks An
  Epidemic Theoretic Framework , poster paper in
  The 3rd IEEE International Conference on
  Distributed Computing in Sensor Systems (DCOSS)
  2007.
• P. De, Y. Liu, and S. K. Das, An Epidemic
  Theoretic Framework for Evaluating Broadcast
  Protocols in Wireless Sensor Networks, In the
  4th IEEE International Conference on Mobile Ad
  Hoc and Sensor Systems (MASS) 2007
• P. De, Y. Liu, and S. K. Das, An Epidemic
  Theoretic Framework for Vulnerability Analysis of
  Broadcast Protocols in Wireless Sensor Networks
  , under review in IEEE Transactions on Mobile
  Computing.
• P. De, Y. Liu, and S. K. Das, Harnessing
  Epidemic Theory to Model Malware Propagation in
  Wireless Sensor Networks, under review in IEEE
  Communications Magazine Special Edition on
  Security in Mobile Ad Hoc and Sensor Networks
• P. De, Y. Liu, and S. K. Das, ReMo An Energy
  Efficient Reprogramming Protocol for Mobile
  sensor Networks, accepted for publication at The
  6th IEEE International Conference on Pervasive
  Computing and Communications (PerCom) 2008.
• Work under preparation
• P. De, Y. Liu, and S. K. Das, An Analytical
  Model for the Performance Analysis of Data
  Dissemination Protocols in Mobile Sensor
  Networks.
• P. De, Y. Liu, and S. K. Das, Analyzing
  Information Propagation over Data Dissemination
  Protocols in Mobile Sensor Networks.