Anomalous Networks
Sevvandi Kandanaarachchi
CSIRO’s Data61
Pint of Science 2023
Work with Rob Hyndman
In a pub we gather round,
Pints in hand, laughter abound,
But talk soon turns to networks strange,
And patterns that we want to change.
Networks, they change and evolve,
Like a story unfolding with time,
Some are ordinary and dull,
But others are juicy and prime.
With drama, intrigue, and political might,
Anomalous networks do excite,
Juicy details, strikes, and scandal,
We want to know more, it's a healthy handle.
Oddnet helps us uncover the truth,
And lays bare the networks, with its sleuth,
It shows us the patterns we couldn't see,
The intrigue and mess in all its glory.
Author: ChatGPT
What is a network from a scientific sense?
We can represent
things/events/relationships
as a network
• Circles are called nodes/vertices
• Lines are called edges/links
• Our network – we give it meaning
In this example
• Nodes are people
• Edges denote if they are friends
• Edge = “is friends with”
Emily
Rose
Ben
Andreas
Kate
Peter
Melbourne train network
• Node = Train station
• Edge = train line
• subnetwork
Burnley
Heyington
Hawthorn
East Richmond
https://en.wikipedia.org/wiki/List_of_railway_stations_in_Melbourne
Actor movie networks
• Node = character
• Link = if the characters appear in
a scene
• Colour of link -> how many times
the characters appear together
https://rockcontent.com/blog/movie-galaxies-uses-social-graph-
organization-to-visualize-movie-interconnectedness/
The Matrix movie network
Computer networks
• Node = physical device
• Edge = connection
LAN: Local Area Network
Smartphones
Printer
Smart Appliances
Data Server
Router
Networks
Static Dynamic
Does not change over time Changes over time
Examples of dynamic networks
• Traffic in computer networks
• Friend/social networks
• Political networks
• Bill voting patterns
• Research collaboration networks
• Virus evolution (protein mutations)
𝑡𝑖𝑚𝑒 = 1 𝑡𝑖𝑚𝑒 = 2 𝑡𝑖𝑚𝑒 = 3 𝑡𝑖𝑚𝑒 = 4
𝑡𝑖𝑚𝑒 = 5
𝑡𝑖𝑚𝑒 = 6
𝑡𝑖𝑚𝑒 = 7 𝑡𝑖𝑚𝑒 = 8
Why are anomalies important?
They tell a different story.
Anomalies can signify important
events!
Examples of anomalies
• Astronomical anomalies – solar flares
• Credit card fraud among billions of legitimate
transactions
• Computer networks - cyber attacks
• Weather anomalies – tsunamis
𝑡𝑖𝑚𝑒 = 1 𝑡𝑖𝑚𝑒 = 2 𝑡𝑖𝑚𝑒 = 3 𝑡𝑖𝑚𝑒 = 4
𝑡𝑖𝑚𝑒 = 5
𝑡𝑖𝑚𝑒 = 6
𝑡𝑖𝑚𝑒 = 7 𝑡𝑖𝑚𝑒 = 8
The transformation
Features
Networks
Features
oddnet
• Once we have features it is much easier
• We can see how features change
• Many statistical methods to model features over time
• Big deviations are anomalous
oddnet
Features
Networks Model features Deviations = anomalous
Why do we need an algorithm if
we can eyeball it?
We need automated algorithms
because . . .
Its not so simple!
• Data arrives very fast
• And lots and lots of data
• Complex data
• Cannot have a person manually finding anomalies
• Example: Recent data breaches
• Applications in cyber security
Example: US Senate Voting patterns
• Why? It is interesting and the data is available.
• Data from the 40th Congress until the 113th Congress
• 40th Congress during 1867 – 1869
• 113th Congress during 2013 – 2015
• Node = senator
• Edge = if they vote on the same bill
• Red = Republican, Blue = Democrat
40th Congress
(Senate)
41st Congress
(Senate)
42nd Congress
(Senate)
43rd Congress
(Senate)
What happened there?
• The Panic of 1873
• Financial crisis, strikes
• Economic depression
Wikipedia says
In the United States, the Panic was known as the
"Great Depression" until the events of 1929 and
the early 1930s set a new standard.
https://pixabay.com/illustrations/financial-crisis-
stock-exchange-544944/
https://www.theage.com.au/national/victoria/protesters-
bring-city-to-a-standstill-for-third-climate-change-protests-
20190920-p52tfw.html
Thank you for listening,
To the stories that we share,
About networks that are anomalous,
And the secrets they ensnare.
- ChatGPT
Thank you!

Anomalous Networks

  • 1.
    Anomalous Networks Sevvandi Kandanaarachchi CSIRO’sData61 Pint of Science 2023 Work with Rob Hyndman
  • 2.
    In a pubwe gather round, Pints in hand, laughter abound, But talk soon turns to networks strange, And patterns that we want to change. Networks, they change and evolve, Like a story unfolding with time, Some are ordinary and dull, But others are juicy and prime. With drama, intrigue, and political might, Anomalous networks do excite, Juicy details, strikes, and scandal, We want to know more, it's a healthy handle. Oddnet helps us uncover the truth, And lays bare the networks, with its sleuth, It shows us the patterns we couldn't see, The intrigue and mess in all its glory. Author: ChatGPT
  • 3.
    What is anetwork from a scientific sense?
  • 4.
  • 5.
    • Circles arecalled nodes/vertices • Lines are called edges/links • Our network – we give it meaning In this example • Nodes are people • Edges denote if they are friends • Edge = “is friends with” Emily Rose Ben Andreas Kate Peter
  • 6.
    Melbourne train network •Node = Train station • Edge = train line • subnetwork Burnley Heyington Hawthorn East Richmond https://en.wikipedia.org/wiki/List_of_railway_stations_in_Melbourne
  • 7.
    Actor movie networks •Node = character • Link = if the characters appear in a scene • Colour of link -> how many times the characters appear together https://rockcontent.com/blog/movie-galaxies-uses-social-graph- organization-to-visualize-movie-interconnectedness/ The Matrix movie network
  • 8.
    Computer networks • Node= physical device • Edge = connection LAN: Local Area Network Smartphones Printer Smart Appliances Data Server Router
  • 9.
    Networks Static Dynamic Does notchange over time Changes over time
  • 10.
    Examples of dynamicnetworks • Traffic in computer networks • Friend/social networks • Political networks • Bill voting patterns • Research collaboration networks • Virus evolution (protein mutations)
  • 11.
    𝑡𝑖𝑚𝑒 = 1𝑡𝑖𝑚𝑒 = 2 𝑡𝑖𝑚𝑒 = 3 𝑡𝑖𝑚𝑒 = 4 𝑡𝑖𝑚𝑒 = 5 𝑡𝑖𝑚𝑒 = 6 𝑡𝑖𝑚𝑒 = 7 𝑡𝑖𝑚𝑒 = 8
  • 12.
  • 13.
    They tell adifferent story. Anomalies can signify important events!
  • 14.
    Examples of anomalies •Astronomical anomalies – solar flares • Credit card fraud among billions of legitimate transactions • Computer networks - cyber attacks • Weather anomalies – tsunamis
  • 15.
    𝑡𝑖𝑚𝑒 = 1𝑡𝑖𝑚𝑒 = 2 𝑡𝑖𝑚𝑒 = 3 𝑡𝑖𝑚𝑒 = 4 𝑡𝑖𝑚𝑒 = 5 𝑡𝑖𝑚𝑒 = 6 𝑡𝑖𝑚𝑒 = 7 𝑡𝑖𝑚𝑒 = 8
  • 16.
  • 17.
  • 18.
    oddnet • Once wehave features it is much easier • We can see how features change • Many statistical methods to model features over time • Big deviations are anomalous
  • 19.
  • 20.
    Why do weneed an algorithm if we can eyeball it?
  • 21.
    We need automatedalgorithms because . . .
  • 22.
    Its not sosimple! • Data arrives very fast • And lots and lots of data • Complex data • Cannot have a person manually finding anomalies • Example: Recent data breaches • Applications in cyber security
  • 23.
    Example: US SenateVoting patterns • Why? It is interesting and the data is available. • Data from the 40th Congress until the 113th Congress • 40th Congress during 1867 – 1869 • 113th Congress during 2013 – 2015 • Node = senator • Edge = if they vote on the same bill • Red = Republican, Blue = Democrat
  • 24.
  • 25.
  • 26.
    What happened there? •The Panic of 1873 • Financial crisis, strikes • Economic depression Wikipedia says In the United States, the Panic was known as the "Great Depression" until the events of 1929 and the early 1930s set a new standard. https://pixabay.com/illustrations/financial-crisis- stock-exchange-544944/ https://www.theage.com.au/national/victoria/protesters- bring-city-to-a-standstill-for-third-climate-change-protests- 20190920-p52tfw.html
  • 27.
    Thank you forlistening, To the stories that we share, About networks that are anomalous, And the secrets they ensnare. - ChatGPT
  • 28.