The Dark Side of Data Mining

1
The Dark Side of Data Mining

• Hannah McLaren
• 100044892

2
Agenda

• What is data mining
• A quick overview of data mining security
• Security breaches
• Ethical issues
• Protecting the integrity of data mining
• Questions

3
Data Mining

• Data mining is applying deductive models to data
  in order to discover patterns
• Whos doing it? Everyone
• AirMiles, Aeroplan
• Banks CIBC, RBC, TD
• Credit card companies
• MSN and Gmail
• The government
• Anyone who markets

4
(No Transcript)
5
Data Mining Security 2 Faces

• The security of the actual data being mined
• Vulnerable to being stolen
• Intentionally incorrect data can also be added
• Unethical companies try to sell their databases
  without consent
• The ethical integrity of the knowledge mined from
  the data
• Can cause bias
• Deliberately misleading inferences

6
Data Mining and E-Commerce

• Earlier presentations have already covered why
  data mining is useful for e-commerce
• Data mining security is a major factor
• To generate accurate models of our customers, we
  have to be confident that the information is
  accurate and valid
• Companies known to lack security or ethics tend
  not to succeed

7
Security Breaches

• How do unwanted miners gain access?
• Guessing easy login/password combinations
• Or, sometimes theres no password protection
• ie Aeroplan
• Brute force attack over the internet
• Access is given as part of a business deal
• the most common method
• Physically stealing the machines
• The information stolen is generally used to
  create fake identities
• Sometimes information is not stolen, but modified

8
Ethical Considerations

• Snooping
• Mining the data with a specific goal in mind
• Searching through the data for a particular
  person/organization
• Dredging
• Imposing patterns on data where none exist
• Or, finding relationships and creating
  interesting explanations
• Elizabeth Taylors marriages and the stock market

9
Ethical Considerations

• Fishing
• Searching for that one key attribute that affects
  most relations
• Not necessarily a valid attribute, could just be
  a fluke
• Reselling
• Offering a database of customer information for
  sale, without customer knowledge or consent

10
Getting a Security Blanket

• Different strategies are being put in place to
  protect data from being stolen or unethically
  used
• Governments are creating stricter standards on
  data collection and mining practices
• Well, some governments.
• Canada has PIPEDA

11
Tech Solutions

• A number of companies are in the business of
  keeping data secure
• RSA offers encryption solutions
• Also a new 2 factor authentication system that
  makes it almost impossible to access an account
  using just a username/pin
• Cisco offers its Adaptive Threat Defence program
  for networks
• Security software relies on defined rules of how
  a legitimate application acts
• ie it knows that a web-based server
  application should not be generating email traffic

12
Tech Solutions

• Mining data ethically is a different problem
• Inductive modelling tools like Cognos Scenario
  are not built to produce amoral models
• Human error, either unintentional or deliberate
  is the cause
• All involved parties need to have knowledge of
  what question is being answered
• Critical analysis of the results
• 95 confidence that the wins/losses in the NFC
  will predict gains/losses on the NSYE does this
  even make sense?
• Using multiple test sets repeatedly to check the
  validity of results

13
In the news

• Microsoft has partnered with the RCMP to create
  CETS (Child Exploitation Tracking System)
• What they needed was a tracking system that
  would not repeat information but link and connect
  criminal behavior online that is difficult for
  the human eye to see.
• John Hancock, senior consultant with Microsoft
  Canada
• CETS will allow police agencies across the
  country to
• Manage and analyze huge volumes of data
• Cross-reference obscure data relationships
• Use social-network analysis to discover
  communities of child predators
• Can they keep this database secure? Ethical?

14
For more information

• RSA http//www.rsasecurity.com/
• Cisco Systems http//www.cisco.com/
• Tool Thwarts Online Child Predators from the
  Microsoft Press sitehttp//channels.microsoft.com
  /presspass/features/2005/apr05/04-07CETS.asp
• Jensen, D. (2000). Data snooping, dredging and
  fishing The dark side of data mining. SIGKDD
  Explorations. 1-2, 52-54.
• Kantardzic, M. (2003). Data mining Concepts,
  models, methods and algorithms. Hoboken, NJ
  Wiley-Interscience, IEEE Press.
• Berry, M. J. A. (1997). Data mining techniques
  for marketing, sales, and customer support. New
  York Wiley.
• Chen, Z. (2002). Intelligent data warehousing
  From data preparation to data mining. Boca Raton,
  FL CRC Press.