Download ICDM-workshop-keynot.. - The University of Texas at Dallas

Data Mining,
Security and Privacy
Prof. Bhavani Thuraisingham
Prof. Murat Kantarcioglu
Ms Li Liu (PhD Student – completing December 2007)
The University of Texas at Dallas
October 28, 2007
Outline
 Data Mining for Security Applications
 Privacy Concerns
 What is Privacy?
 Why is data mining a threat to privacy
 Developments in Privacy
 Directions for Privacy
Data Mining Needs for Counterterrorism:
Non-real-time Data Mining
 Gather data from multiple sources
- Information on terrorist attacks: who, what, where, when, how
- Personal and business data: place of birth, ethnic origin,
religion, education, work history, finances, criminal record,
relatives, friends and associates, travel history, . . .
- Unstructured data: newspaper articles, video clips, speeches,
emails, phone records, . . .
 Integrate the data, build warehouses and federations
 Develop profiles of terrorists, activities/threats
 Mine the data to extract patterns of potential terrorists and predict
future activities and targets
 Find the “needle in the haystack” - suspicious needles?
 Data integrity is important
 Techniques have to SCALE
Data Mining Needs for Counterterrorism:
Real-time Data Mining
 Nature of data
- Data arriving from sensors and other devices

Continuous data streams
- Breaking news, video releases, satellite images
- Some critical data may also reside in caches
 Rapidly sift through the data and discard unwanted data for later use
and analysis (non-real-time data mining)
 Data mining techniques need to meet timing constraints
 Quality of service (QoS) tradeoffs among timeliness, precision and
accuracy
 Presentation of results, visualization, real-time alerts and triggers
Data Mining for Real-time Threats
Integrate
data
sources in
real-time
Rapidly
sift through
data and
discard
irrelevant
data
Build
real-time
models
Mine
the
data
Data sources
with information
about terrorists
and terrorist activities
Report
final
results
Examine
Results in
Real-time
What should be done: Form a Research Agenda
 Immediate action (0 - 1 year)
- We’ve got to know what our current capabilities are
- Do the commercial tools scale? Do they work only on special
data and limited cases? Do they deliver what they promise?
- Need an unbiased objective study with demonstrations
 At the same time, work on the big picture
- What do we want? What are our end results for the foreseeable
future? What are the criteria for success? How do we evaluate
the data mining algorithms? What testbeds do we build?
 Near-term (1 - 3 years)
- Leverage current efforts
- Fill the gaps in a goal-directed way; technology transfer
 Long-term (3 - 5 years and beyond)
- 5-year
R&D plan for data mining for counterterrorism
IN SUMMARY:
 Data Mining is very useful to solve Security Problems
- Data mining tools could be used to examine audit data
-
-
and flag abnormal behavior
Much recent work in Intrusion detection (unit #18)
 e.g., Neural networks to detect abnormal patterns
Tools are being examined to determine abnormal patterns
for national security
 Classification techniques, Link analysis
Fraud detection
 Credit cards, calling cards, identity theft etc.
BUT CONCERNS FOR PRIVACY
What is Privacy
 Medical Community
- Privacy is about a patient determining what information
the doctor should release about him/her
 Financial community
- A bank customer determines what financial information
the bank should release about him/her
 Government community
- FBI would collect information about US citizens. However
FBI determines what information about a US citizen it can
release to say the CIA
Some Privacy concerns
 Medical and Healthcare
- Employers, marketers, or others knowing of private medical
concerns
 Security
- Allowing access to individual’s travel and spending data
- Allowing access to web surfing behavior
 Marketing, Sales, and Finance
- Allowing access to individual’s purchases
Data Mining as a Threat to Privacy
 Data mining gives us “facts” that are not obvious to human analysts
of the data
 Can general trends across individuals be determined without
revealing information about individuals?
 Possible threats:
Combine collections of data and infer information that is private
 Disease information from prescription data
 Military Action from Pizza delivery to pentagon
 Need to protect the associations and correlations between the data
that are sensitive or private
-
Some Privacy Problems and Potential Solutions
 Problem: Privacy violations that result due to data mining
- Potential solution: Privacy-preserving data mining
 Problem: Privacy violations that result due to the Inference problem
- Inference is the process of deducing sensitive information from
the legitimate responses received to user queries
- Potential solution: Privacy Constraint Processing
 Problem: Privacy violations due to un-encrypted data
- Potential solution: Encryption at different levels
 Problem: Privacy violation due to poor system design
- Potential solution: Develop methodology for designing privacyenhanced systems
Privacy as Inference:
Privacy Constraint Processing
 Privacy constraint/policy processing
- Based on prior research in security constraint processing
- Simple Constraint: an attribute of a document is private
- Content-based constraint: If document contains information
about X, then it is private
- Association-based Constraint: Two or more documents taken
together is private; individually each document is public
- Release constraint: After X is released Y becomes private
 Augment a database system with a privacy controller for constraint
processing
Architecture for Privacy
Constraint Processing
User Interface Manager
Privacy
Constraints
Constraint
Manager
Query Processor:
Constraints during
query and release
operations
DBMS
Database Design
Tool
Update
Processor:
Constraints during
database design
operation
Constraints
during update
operation
Database
Semantic Model for Privacy Control
Dark lines/boxes contain
private information
Cancer
Influenza
Has disease
John’s
address
Patient John
address
England
Travels frequently
Privacy Preserving Data Mining
 Prevent useful results from mining
- Introduce “cover stories” to give “false” results
- Only make a sample of data available so that an adversary is
unable to come up with useful rules and predictive functions
 Randomization
- Introduce random values into the data and/or results
- Challenge is to introduce random values without significantly
affecting the data mining results
- Give range of values for results instead of exact values
 Secure Multi-party Computation
- Each party knows its own inputs; encryption techniques used to
compute final results
Cryptographic Approaches for
Privacy Preserving Data Mining
 Secure Multi-part Computation (SMC) for PPDM
- Mainly used for distributed data mining.; Provably secure under
some assumptions.; Learned models are accurate; Mainly semihonest assumption (i.e. parties follow the protocols); Malicious
model is also explored recently. (e.g. Kantarcioglu and Kardes
paper in this workshop); Many SMC based PPDM algorithms share
common sub-protocols (e.g. dot product, summation, etc. )
 Drawbacks:
- Still not efficient enough for very large datasets. (e.g. petabyte sized
datasets ??); Semi-honest model may not be realistic;Malicious
model is even slower
 Possible new directions
- New models that can trade-off better between efficiency and
security; Game theoretic / incentive issues in PPDM; Combining
anonymization and cryptographic techniques for PPDM
Perturbation Based Approaches for
Privacy Preserving Data Mining
 Goal: Distort data while still preserve some properties for data
mining propose.
− Additive Based
− Multiplicative Based
− Condensation based
− Decomposition
− Data Swapping
 Goal: Achieve a high data mining accuracy with maximum
privacy protection.
 Our approach: Privacy is a personal choice, so should be
individually adaptable
(Liu, Kantarcioglu and Thuraisingham ICDM’06)
Perturbation Based Approaches for
Privacy Preserving Data Mining
 The trend is to make PPDM approaches to reflect reality
 We investigated perturbation based approaches with real-
world data sets
 We give a applicability study to the current approaches
Liu, Kantarcioglu and Thuraisingham, DKE 07
 We found out,
- The reconstruction of the original distribution may not
work well with real-world data sets
Try to modify perturbation techniques, and adapt some
data mining tools, e.g. Liu, Kantarcioglu and
Thuraisingham, Novel decision tree – UTD technical report 06
-
CPT: Confidentiality, Privacy and Trust
 I as a user of Organization A send data about me to
organization B, I read the privacy policies enforced by
organization B
- If I agree to the privacy policies of organization B, then I
will send data about me to organization B
- If I do not agree with the policies of organization B, then I
can negotiate with organization B
 Even if the web site states that it will not share private
information with others, do I trust the web site
 Note: while confidentiality is enforced by the organization,
privacy is determined by the user. Therefore for
confidentiality, the organization will determine whether a user
can have the data. If so, then the organization van further
determine whether the user can be trusted
Platform for Privacy Preferences (P3P):
What is it?
 P3P is an emerging industry standard that enables web sites
t9o express their privacy practices in a standard format
 The format of the policies can be automatically retrieved and
understood by user agents
 It is a product of W3C; World wide web consortium
www.w3c.org
 When a user enters a web site, the privacy policies of the web
site is conveyed to the user; If the privacy policies are
different from user preferences, the user is notified; User can
then decide how to proceed
 Several major corporations are working on P3P standards
including
Privacy for Assured Information Sharing
Data/Policy for Federation
Export
Data/Policy
Export
Data/Policy
Export
Data/Policy
Component
Data/Policy for
Agency A
Component
Data/Policy for
Agency C
Component
Data/Policy for
Agency B
Privacy Preserving Surveillance
Raw video surveillance data
Face Detection and
Face
Derecognizing
system
Faces of trusted
people derecognized
to preserve privacy
Suspicious Event
Detection System
Manual Inspection
of video data
Suspicious people
found
Suspicious events
found
Report of security personnel
Comprehensive
security report
listing suspicious
events and people
detected
Directions: Foundations of Privacy Preserving
Data Mining
 We proved in 1990 that the inference problem in
general was unsolvable, therefore the suggestion
was to explore the solvability aspects of the
problem.
 Can we do something similar for privacy?
Is the general privacy problem solvable?
What are the complicity classes?
What is the storage and time complicity
 We need to explore the foundation of PPDM and
related privacy solutions
-
Directions: Testbed Development and Application
Scenarios
 There are numerous PPDM related algorithms. How
do they compare with each other? We need a
testbed with realistic parameters to test the
algorithms
 It is time to develop real world scenarios where
these algorithms can be utilized
 Is it feasible to develop realistic commercial
products or should each organization adapt
product to suit their needs?
Key Points
 1. There is no universal definition for privacy, each
organization must definite what it means by privacy and
develop appropriate privacy policies
 2. Technology alone is not sufficient for privacy We need
technologists, Policy expert, Legal experts and Social
scientists to work on Privacy
 3. Some well known people have said ‘Forget about privacy”
Therefore, should we pursue research on Privacy?
- Interesting research problems, there need to continue
with research
- Something is better than nothing
- Try to prevent privacy violations and if violations occur
then prosecute
 4. We need to tackle privacy from all directions
Application Specific Privacy?
 Examining privacy may make sense for healthcare and
financial applications
 Does privacy work for Defense and Intelligence applications?
 Is it even meaningful to have privacy for surveillance and
geospatial applications
- Once the image of my house is on Google Earth, then how
much privacy can I have?
I may want my location to be private, but does it make
sense if a camera can capture a picture of me?
- If there are sensors all over the place, is it meaningful to
have privacy preserving surveillance?
 This suggests that we need application specific privacy
 It is not meaningful to examine PPDM for every data mining
algorithm and for every application
-
Data Mining and Privacy: Friends or Foes?
 They are neither friends nor foes
 Need advances in both data mining and privacy
 Need to design flexible systems
- For some applications one may have to focus entirely on
“pure” data mining while for some others there may be a
need for “privacy-preserving” data mining
- Need flexible data mining techniques that can adapt to the
changing environments
 Technologists, legal specialists, social scientists, policy
makers and privacy advocates MUST work together