Collection of general data mining briefings

1
Social Networking, Security and Privacy
Dr. Bhavani Thuraisingham
April 21, 2008

2
Social Networkshttp//www.flairandsquare.com/arch
ives/167

• A social network site allows people who share
  interests to build a trusted network/ online
  community. A social network site will usually
  provide various ways for users to interact, such
  as IM (chat/ instant messaging), email, video
  sharing, file sharing, blogging, discussion
  groups, etc.
• The main types of social networking sites have a
  theme, they allow users to connect through
  image or video collections online (like Flicker
  or You Tube) or music (like My Space, lastfm).
  Most contain libraries/ directories of some
  categories, such as former classmates, old work
  colleagues, and so on (like Face book, friends
  reunited, Linked in, etc). They provide a means
  to connect with friends (by allowing users to
  create a detailed profile page), and recommender
  systems linked to trust.

3
Popular Social Networks

• Face book - A social networking website.
  Initially the membership was restricted to
  students of Harvard University. It was originally
  based on what first-year students were given
  called the face book which was a way to get to
  know other students on campus. As of July 2007,
  there over 34 million active members worldwide.
  From September 2006 to September 2007 it
  increased its ranking from 60 to 6th most visited
  web site, and was the number one site for photos
  in the United States.
• Twitter- A free social networking and
  micro-blogging service that allows users to send
  updates (text-based posts, up to 140 characters
  long) via SMS, instant messaging, email, to the
  Twitter website, or an application/ widget within
  a space of your choice, like MySpace, Facebook, a
  blog, an RSS Aggregator/reader.
• My Space - A popular social networking website
  offering an interactive, user-submitted network
  of friends, personal profiles, blogs, groups,
  photos, music and videos internationally.
  According to AlexaInternet, MySpace is currently
  the worlds sixth most popular English-language
  website and the sixth most popular website in any
  language, and the third most popular website in
  the United States, though it has topped the chart
  on various weeks. As of September 7, 2007, there
  are over 200 million accounts.

4
Social Networks More formal definition

• A structural approach to understanding social
  interaction.
• Networks consist of Actors and the Ties between
  them.
• We represent social networks as graphs whose
  vertices are the actors and whose edges are the
  ties.
• Edges are usually weighted to show the strength
  of the tie.
• In the simplest networks, an Actor is an
  individual person.
• A tie might be is acquainted with. Or it might
  represent the amount of email exchanged between
  persons A and B.

5
Social Network Examples

• Effects of urbanization on individual well-being
• World political and economic system
• Community elite decision-making
• Social support, Group problem solving
• Diffusion and adoption of innovations
• Belief systems, Social influence
• Markets, Sociology of science
• Exchange and power
• Email, Instant messaging, Newsgroups
• Co-authorship, Citation, Co-citation
• SocNet software, Friendster
• Blogs and diaries, Blog quotes and links

6
History

• Sociograms were invented in 1933 by Moreno.
• In a sociogram, the actors are represented as
  points in a two-dimensional space. The location
  of each actor is significant. E.g. a central
  actor is plotted in the center, and others are
  placed in concentric rings according to
  distance from this actor.
• Actors are joined with lines representing ties,
  as in a social network. In other words a social
  network is a graph, and a sociogram is a
  particular 2D embedding of it.
• These days, sociograms are rarely used (most
  examples on the web are not sociograms at all,
  but networks). But methods like MDS
  (Multi-Dimensional Scaling) can be used to lay
  out Actors, given a vector of attributes about
  them.
• Social Networks were studied early by researchers
  in graph theory (Harary et al. 1950s). Some
  social network properties can be computed
  directly from the graph.
• Others depend on an adjacency matrix
  representation (Actors index rows and columns of
  a matrix, matrix elements represent the tie
  strength between them).

7
Social Networks Basic Questions

• Balance important in exchange networks
• In a two-person network (dyad), exchange of
  goods, services and cash should be balanced.
• More generally, exchanges of favors or
  support are likely to be quite balanced.
• Role what role does the actor perform in the
  network?
• Role is defined in terms of Actors
  neighborhoods.
• The neighborhood is the set of ties and actors
  connected directly to the current actor.
• Actors with similar or identical neighborhoods
  are assigned the same role.
• What is the related idea from semiotics?
• Paradigm interchangability. Actors with the
  same role areinterchangable in the network.

8
Social Networks Basic Questions

• Prestige How important is the actor in the
  network?
• Related notions are status and centrality.
• Centrality reifies the notion of peripheral vs.
  central participation from communities of
  practice.
• Key notions of centrality were developed in the
  1970s, e.g. eigenvalue centrality by Bonacich.
• Most of these measures were rediscovered as
  quality measures for web pages
• Indegree
• Pagerank eigenvalue centrality
• HITS ? two-mode eigenvalue centrality

9
Social Network Concepts

• Actor
• An actor is a basic component for SNs. Actors
  can be
• Individual people, Corporations, Nation-States,
  Social groups
• Modes
• If all the actors are of the same type, the
  network is called a one-mode network. If there
  are two groups of actor then it is a two-mode
  network.
• E.g. an affiliation network is a two-mode
  network. One mode is individuals, the other is
  groups to which they belong. Ties represent the
  relation person A is a member of group B.
• Ties
• A tie is the relation between two actors. Common
  types of ties include
• Friendship, Amount of communication, Goods
  exchanged, Familial relation (kinship),
  Institutional relations

10
Practical issues Boundaries and Samples

• Because human relations are rich and unbounded,
  drawing meaningful boundaries for network
  analysis is a challenge.
• There are two main approaches
• Realist boundaries perceived by actors
  themselves, e.g. gang members or ACM members.
• Nominalist Boundaries created by researcher
  e.g. people who publish in ACM CHI.
• To deal with large networks, sampling is
  necessary. Unfortunately, randomly sampled graphs
  will typically have completely different
  structure. Why?
• One approach to this is snowballing. You start
  with a random sample. Then extend with all actors
  connected by a tie. Then extend with all actors
  connected to the previous set by a tie

11
Social Network Analysis of 9/11 Terrorists
(www.orgnet.com)
Early in 2000, the CIA was informed of two
terrorist suspects linked to al-Qaeda. Nawaf
Alhazmi and Khalid Almihdhar were photographed
attending a meeting of known terrorists in
Malaysia. After the meeting they returned to Los
Angeles, where they had already set up
residence in late 1999.
12
Social Network Analysis of 9/11 Terrorists

• What do you do with these suspects? Arrest or
  deport them immediately? No, we need to use them
  to discover more of the al-Qaeda network.
• Once suspects have been discovered, we can use
  their daily activities to uncloak their network.
  Just like they used our technology against us, we
  can use their planning process against them.
  Watch them, and listen to their conversations to
  see...
• who they call / email
• who visits with them locally and in other cities
• where their money comes from
• The structure of their extended network begins to
  emerge as data is discovered via surveillance.

13
Social Network Analysis of 9/11 Terrorists
A suspect being monitored may have many contacts
-- both accidental and intentional. We must
always be wary of 'guilt by association'.
Accidental contacts, like the mail delivery
person, the grocery store clerk, and neighbor may
not be viewed with investigative interest.
Intentional contacts are like the late
afternoon visitor, whose car license plate is
traced back to a rental company at the airport,
where we discover he arrived from Toronto (got to
notify the Canadians) and his name matches a cell
phone number (with a Buffalo, NY area code) that
our suspect calls regularly. This intentional
contact is added to our map and we start tracking
his interactions -- where do they lead? As data
comes in, a picture of the terrorist organization
slowly comes into focus. How do investigators
know whether they are on to something big? Often
they don't. Yet in this case there was another
strong clue that Alhazmi and Almihdhar were up to
no good -- the attack on the USS Cole in October
of 2000. One of the chief suspects in the Cole
bombing Khallad was also present along with
Alhazmi and Almihdhar at the terrorist meeting
in Malaysia in January 2000. Once we have their
direct links, the next step is to find their
indirect ties -- the 'connections of their
connections'. Discovering the nodes and links
within two steps of the suspects usually starts
to reveal much about their network. Key
individuals in the local network begin to stand
out. In viewing the network map in Figure 2, most
of us will focus on Mohammed Atta because we now
know his history. The investigator uncloaking
this network would not be aware of Atta's
eventual importance. At this point he is just
another node to be investigated.
14
Social Network Analysis of 9/11 Terrorists
Figure 2 shows the two suspects and
15
Social Network Analysis of 9/11 Terrorists
Figure 2 shows the two suspects and
Atta's eventual importance. At this point he is
just another node to be investigated.
                                                  
                                                  
                                                  
                                                  
         Figure 3 shows the direct
16
Social Network Analysis of 9/11 Terrorists

• We now have enough data for two key conclusions
• All 19 hijackers were within 2 steps of the two
  original suspects uncovered in 2000!
• Social network metrics reveal Mohammed Atta
  emerging as the local leader
• With hindsight, we have now mapped enough of the
  9-11 conspiracy to stop it. Again, the
  investigators are never sure they have uncovered
  enough information while they are in the process
  of uncloaking the covert organization. They also
  have to contend with superfluous data. This data
  was gathered after the event, so the
  investigators knew exactly what to look for.
  Before an event it is not so easy.
• As the network structure emerges, a key dynamic
  that needs to be closely monitored is the
  activity within the network. Network activity
  spikes when a planned event approaches. Is there
  an increase of flow across known links? Are new
  links rapidly emerging between known nodes? Are
  money flows suddenly going in the opposite
  direction? When activity reaches a certain
  pattern and threshold, it is time to stop
  monitoring the network, and time to start
  removing nodes.
• The author argues that this bottom-up approach of
  uncloaking a network is more effective than a top
  down search for the terrorist needle in the
  public haystack -- and it is less invasive of the
  general population, resulting in far fewer "false
  positives".

17
Social Network Analysis of Steroid Usage in
Baseball (www.orgnet.com)
Figure 2 shows the two suspects and
When the Mitchell Report on steroid use in Major
League Baseball MLB, was published, people were
surprised at who and how many players were
mentioned. The diagram below shows a human
network created from data found in the Mitchell
Report. Baseball players are shown as green
nodes. Those who were found to be providers of
steroids and other illegal performance enhancing
substances appear as red nodes. The links reveal
the flow of chemicals -- from provider to player.
18
Knowledge Management Examples

• Managing the 21st Century Organization
• Networks of Adaptive/Agile Organizations
• Best Practice Organizational Network Mapping
• Discovering Communities of Practice
• Data-Mining E-mail
• Finding Leaders on your Team
• Post-Merger Integration
• Knowledge Sharing in Organizations
• Innovation happens at the Intersections
• Partnerships and Alliances in Industry
• Decision-Making in Organizations
• New Organizational Structures

19
Knowledge Sharing in Organizations Finding
Experts

Figure 2 shows the two suspects and
20
Knowledge Sharing Network Finding Experts
(www.orgnet.com)

Figure 2 shows the two suspects and
Organizational leaders are preparing for the
potential loss of expertise and knowledge flow
due to turnover, downsizing, outsourcing, and the
coming retirements of the baby boom generation.
The model network (previous chart) is used to
illustrate the knowledge continuity analysis
process. Each node in this sample network
(previous chart) represents a person that works
in a knowledge domain. Some people have more /
different knowledge than others. Employees who
will retire in 2 years or less have their nodes
colored red. Those who will retire in 3-4 years
are colored yellow. Those retiring in 5 years or
later are colored green. A gray, directed line
is drawn from the seeker of knowledge to the
source of expertise. A--B indicates that A seeks
expertise / advice from B. Those with many
arrows pointing to them are sought often for
assistance. The top subject matter experts --
SMEs -- in this group are nodes 29, 46, 100, 41,
36 and 55. The SMEs were discovered using a
network metric in InFlow that is similar to how
the Google search engine ranks web pages --
using both direct and indirect links. Of the top
six SMEs in this group, half are colored red100
or yellow46, 55. The loss of person 46 has the
greatest potential for knowledge loss. 90 of the
network is within 3 steps of accessing this key
knowledge source.
21
Social Networks Security and Privacy Issues
European Network and Information Security Agency

• The European Network and Information Security
  Agency (ENISA) has released its first issue paper
  Security Issues and Recomendations for Online
  Social Networks".
• http//www.enisa.europa.eu/doc/pdf/deliverables/en
  isa_pp_social_networks.pdf
• Four groups of threats privacy related threats,
  variants of traditional network and information
  security threats, identity related threats,
  social threats.
• Recommendations are given for governments
  (oversight and adaption of existing data
  protection legislation), companies that run such
  networks, technology developers, and research and
  standardisation bodies.
• Some concenrs recommnendation to use automated
  filters against "offensive, litigious or illegal
  content". This brings potential freedom of speech
  issues. European Digital Rights has started a
  campaign against a similar recommendation by the
  Council of Europe.Issue of portability of
  profiles social graphs are also addressed.
  However what is missing is that Information
  about social links is not about only one user,
  but also the others which he is linked to. They
  have to agree if this information is moved to
  different platforms.

22
Social Networks Security and Privacy Issues
Microsoft Recommendations http//www.microsoft.com
/protect/yourself/personal/communities.mspx

• Online communities require you to provide
  personal information. Profiles are public.
  Comments you post are permanently recorded on the
  community site.You might even mention when you
  plan to be out of town.
• E-mail and phishing scammers count on the
  appealing sense of trust that is often fostered
  in online communities to steal your personal
  information. The more you reveal in profiles and
  posts, the more vulnerable you are to scams,
  spam, and identity theft.
• Here are some features to look for when you're
  considering joining an online community
• Privacy policies that explain exactly what
  information the service will collect and how it
  might be used. User guidelines that outline a
  basic code of conduct for users on their sites.
  Sites have the option to penalize reported
  violators with account suspension or
  termination.Special provisions for children and
  their parents, such as family-friendly options
  geared towards protecting children under a
  certain age.Password protection to help keep
  your account secure..E-mail address hiding,
  which lets you display only part of your e-mail
  address on the site's membership lists. Filtering
  options Offered on blogging sites, these tools
  let you to choose which subscribers can see what
  you've written.

23
Appendix Social Networks and Surveillance
Evaluating Suspicion by Association
Ryan Layfield, PhD Student Prof. Bhavani
Thuraisingham
September 2006

24
Overview

• Introduction
• Our Goal
• System Design
• Social Networks
• Threat Detection
• Correlation Analysis
• The Experiment
• Setup
• Current Results
• Issues
• Future Work
• Directions

25
Introduction

• Automated message surveillance is essential to
  communication monitoring
• Widespread use of electronic communication
• Exponential data growth
• Impossible to sift through all by hand
• Going beyond basic surveillance
• Identifying groups rather than individuals
• Monitoring conversations rather than messages

26
Our Approach

• Design new techniques and apply existing
  algorithms to
• Create a machine-understandable model of existing
  social networks Identify abnormal conversations
  and behavior Monitor a given communications
  system in real-time Continuously learn and adapt
  to a dynamic environment
• System Design Three major components
• Social Network Modeler Initial Activity
  Detector Correlated Activity Investigator
• Assumptions
• Individuals engaged in suspicious or undesirable
  behavior rarely act alone
• We can infer than those associated with a person
  positively identified as suspicious have a high
  probability of being either
• Accomplices (participants in suspicious activity)
• Witnesses (observers of suspicious activity)
• Making these assumptions, we create a context of
  association between users of a communication
  network

27
Social Networks

• Within our model
• Every node is a unique user
• Every message creates or strengthens a link
  between nodes
• Over time, the network changes
• Frequent communication leads to stronger links
• Intermittent messaging implies weakening social
  ties
• The strength of the link implies how strong an
  association between individuals is
• From this data, we can theoretically identify
• Hubs
• Groups
• Liaisons

28
Social Networks
29
Threat Detection and Correlation Analysis

• Every message sent is scrutinized in the interest
  of identifying suspicious communication
• Keywords analysis Prior context (i.e. previous
  message content)
• When a detection algorithm yields a strong
  result, a token is created
• The token is created at the origin and passed to
  the recipients) Existing tokens, if any, are
  cloned instead
• The result is a web that potentially reflects the
  dissemination of suspicious information activity
• Future messages with similar suspicious topics
  are not always identifiable with the same
  initial techniques
• Quick replies Pronoun use Assumption that
  recipient is aware of topic
• If a token is present at the sender when a
  message is sent
• Message token is associated with and new message
  are analyzed If analysis yields a strong match,
  the token is further cloned and passed to
  recipient

30
The Experiment

• A rare set of words shared between two or more
  messages are candidates for keyword analysis, but
  they are not always easily sifted from noise
• Noise within text-based messages comes in a
  variety of forms
• Misspelled words
• Unusual word choice
• Incompatible variations of the same language
  (i.e. British vs. American English)
• Unexpected language
• However, we do not want to eliminate potential
  keywords
• Document names
• Terminology specific to a subject
• Buzz words
• We proposed an experiment that attempts to
  eliminate false positives due to noisy data while
  strengthening and expanding our correlation
  techniques

31
Setup

• Tools
• Running word rank database
• Implementation of word set theory infrastructure
• JAMA Matrix Library
• Singular Value Decomposition
• Our Approach
• Apply SVD noise filtering based on 100 messages
• Analyze word frequency correlation between
  current message and prior suspicious messages
• Generate a score based on the results

32
Setup

• Construct a matrix based on the last 100 messages

messages
More common
words
Less common
33
Setup

• Decompose and rebuild

VT
?
U
A
Eliminate weak singular values
34
Setup
Pulled from messages j and k
Raw total score for word wi
Pulled from running word database
Counts only intersection of words
Predefined fixed threshold
35
Current Results
Method is not currently accurate Large
fluctuations Correlation easily swayed by
plethora of common words Uncommon words not given
enough weight
1000 messages evaluated, first 100 used to seed
word ranks.
36
Issues and Directions

• Word frequencies fluctuate wildly during
  beginning of experiment (0.0 10.0)
• Extreme cost for current construction methods and
  computation
• Filtering context limited to recent global
  history
• Affected by large bodies of text
• Future Directions include
• Tap potential of existing matrix for further
  analysis
• Adaptive filtering feedback algorithms
• Speed improvements to accommodate real-time
  streams
• Flexible communication platform monitoring
• Addition of pipe architecture for modular threat
  detection and correlation