Panagiotis ANGELOUDIS,

1
SECURITY AND RELIABILITY OF THE LINER
CONTAINER-SHIPPING NETWORK ANALYSIS OF
ROBUSTNESS USING A COMPLEX NETWORK FRAMEWORK

• Panagiotis ANGELOUDIS,
• Khalid BICHOU, Michael G H BELL,
• David FISK
• Port Operations Research and Technology Centre
• Centre for Transport Studies
• Imperial College London

P O R T e C
2
Theoretical Background
P O R T e C

• Application of complex network theory
• Fast growing field of applied mathematics
• Closely related to graph theory
• Examination of
• Network architecture,
• Error and attack robustness,
• Node degree distributions, etc
• Interesting results in diverse fields
• Eg ecology, sociology, computer science,
  transportation, genetics
• Two main models Small-worlds (Watts
  Strogatz)
• Scale-free network (Barabasi Albert)
• Both developed in late 90s
• Enormous drive among researchers to classify real
  world systems to either of the two models (unique
  properties)

3
Introduction to small worlds
P O R T e C

• Two fundamental network types Regular and Random
  graphs
• Small world occupy the spectrum between
• With varying degrees of randomness
• Name origins Sociology
• Well known experiment by social psychologist
  Stephen Milgram
• Six degrees of separation in the US population
• True distance between two
• random (and seemingly distant)
• network nodes significantly
• smaller than expected

4
Introduction to scale-free networks
P O R T e C

• More relevant to our study
• Contain handful of nodes with exceptionallylarge
  number of links
• Name origins? Links not evenly distributed
  among nodes? Nodes in same network may have few
  hundreds or thousands!? Link distribution
  lacks uniform scale? Scale free
• Network evolution by preferential attachment
  (rich get richer)
• New nodes benefit from their increased
  connectivity
• Study of network robustness, two distinct failure
  modes
• Errors (random node removal) ? Very resilient
• Attacks (planned node removal) ? Exceptionally
  vulnerable
• Example
  ?

5
Scale-free network failure
P O R T e C
6
Current Situation
P O R T e C

• Previous transport networks examined
• Air travel Grid
• USA highway networks
• National rail networks (India)
• Urban rail networks (London Tube, New York
  Subway, etc)
• Some application to supply chains, but not
  maritime settings
• Clear network patterns
• Three major trade lanes (trans-pacific,
  trans-atlatic, trans-asia)
• Evolving global containership network
• Not through strategic master planning and network
  study
• But from individual and localized micro decisions

7
Focus on security
P O R T e C

• Post 9/11
• High priority research robustness and
  survivability of maritime network against nodal
  failures (land / sea attacks, failures, strikes)
• Research has looked at such topics, but has also
  been fragmented into different areas
• System vulnerability, Risk attack avoidance,
  mitigation strategies
• Current approach logical mapping of internal
  processes (ISPS)
• No applied research on the collective robustness
  of the network
• Need to look at the big picture
• Robustness and reliability against
  random/targeted failures taken for granted.
• Not just malicious/unexpected actions like
  terrorism and piracy
• But also union strikes, ship collision, safety
  incidents, weather conditions, IT system failure
  (particularly on automated terminals)

8
Development of a network model
P O R T e C

• Using concepts mentioned above
• Model subset of global shipping network
• Custom maritime traffic simulator
• At this time focus on routes linking European
  N. American ports
• - Network formed as a
• collection of scheduled
• routes
• - We essentially examine
• the flow of containers
• between ports

9
Model parameters
P O R T e C

• Most Trans-Atlantic routes are part of the wider
  global network (eg. round-the world trips) they
  were included in full
• Resulting model has 159 nodes 1558 links
• Complex network model properties not fully
  developed at such small network sizes
• Network distribution strongly suggestive of a
  scale free configuration (consequences of such
  configuration well known and discussed above)

10
Model parameters
P O R T e C

• Many more parameters and statistics!
• Average journey between two random ports 6 stops
• Longest journey between two ports 28 stops
• Network hub identification (vulnerability)
• Ports with largest
• Number of directly linked neighbors
• Number of incoming outgoing services
• Path Optimality Index (POI)(number of optimum
  paths between any two ports that an examined
  port is a part of)

11
Vulnerability Analysis
P O R T e C

• Having identified network hubs, we simulated
  informed attacks
• Impact analysis
• Developed algorithms for automatic container
  rerouting
• New points of transshipment minimum cost paths
• Avoidance of defective network portions
• Estimation of changes in transshipment load borne
  by ports
• Before after
• - Failure in Singapore (red)
• - Shades of blue increasing
• transshipment load
• - Global effects (West Coast,
• Far East, Europe etc)
• - Quantification of impact

12
Future Plans

• More analysis
• Further understanding of structure, network
  properties and robustness
• Elimination of current assumptions
• Completion of maritime route database
• Timetables taken into account
• Modeling of port parameters
• Port capacity!
• New failure mode Port capacity overflow
• Cascading failure
• More statistics!
• Advanced network visualization

PORTeC
13
Thank you!
PORTeC