Into the Data Mine

1
Into the (Data) Mine!
GENERAL
Corruption Prevention Network
14 September 2006
SEGMENT
AUDIENCE
DATE
Presented byAlicia Peters Fraud Prevention and
Control Australian Taxation Office
www.ato.gov.au
2
Organisation
3
What is data mining?

• The nontrivial extraction of implicit,
  previously unknown, and potentially useful
  information from data.
• analysing data to show patterns or
  relationships sorting through large amounts of
  data and picking out pieces of relative
  information or patterns that occur
• http//en.wikipedia.org/w/index.php?titleData_min
  ingoldid59371931

4
Data mining tools

• There are many contemporary tools available
• Tax Office
• Microsoft Excel
• Microsoft Access
• ACL
• Win IDEA

5
The data

• Who owns it?
• Internal
• External
• Where is it?
• Quality
• Duplicates
• Inconsistencies
• Incomplete

6
What data mining can do

• Compare different sets of data
• Identify trends and outliers
• Identify relationships

7
What data mining cant do

• Data mining cannot
• identify fraud in isolation generally requires
  further scrutiny
• verify individual transactions
• avoid false positives and / or false negatives
• work without access to data
• Data mining is less effective if the quality and
  quantity of available data is reduced

8
Outcomes

• Data mining needs to be directed by the type of
  outcome sought
• Identification of potential fraud incidents
• Identification of fraud risks
• Identification of system / process vulnerabilities

9
Where do the ideas / projects come from?

• Criminals perspective of our systems and
  processes
• Communication between Internal Investigations,
  Fraud Control Planning and Official Conduct Team
• External information from liaison with other
  departments
• Research
• Media

10
Planning data mining projects

• Projects are based on emerging risks
• No strict work plan - maintains flexibility to
  deliver products and address emerging risks
• Connections with Internal Audit

11
What we do with the results

• Suspected fraud incidents
• Internal fraud investigation referrals
• Fraud risks and vulnerabilities
• System / process owners
• Fraud Control Planning as an intelligence product
• Tax Office Executive

12
Products

• Threat assessments
• Data mining reports
• Fraud investigation referrals
• Tax Office Executive reports
• Weekly staff newsletter articles

13
Data mining in the Tax Office from an internal
fraud perspective

• Two types are undertaken
• Departmental What makes the department work
• Administered What the department exists for

14
Data mining departmental systems

• 22,000 staff
• 60 locations around Australia
• Projects
• Cabcharge
• Mobile telephones

15
Cabcharge

• Context
• Cabcharge cards and vouchers
• 135,500 transactions a year

16
Cabcharge

• Data
• Internal data quality was poor because of
  different management systems
• Data sought directly from Cabcharge
• Many duplicates
• Incomplete data population
• Inaccuracies around pick up and destination
  locations
• Differences in data for card and voucher
  transactions
• Imported 315,500 transactions into IDEA
• Got rid of duplicates

17
Cabcharge

• Focus
• Large single expenses on cabs
• Use of Cabcharge cards on Saturday

18
Cabcharge

• Validation process
• Large fares
• 35 emails sent to managers to confirm expenditure
• Weekend use
• 301 emails sent to managers to confirm expenditure

19
Cabcharge

• Results
• Fraud
• Suspected personal travel
• Fraud risks
• Incorrect amounts approved
• Observations
• High traffic routes

20
Cabcharge

• Reporting
• Tax Office Executive
• Process owners
• Article in weekly staff newsletter

21
Mobile telephones

• Liaison meeting with Defence
• Identified personal use of mobile telephones as
  an emerging risk especially in relation to
  190 numbers
• Identified that these types of calls were not
  being checked

22
Mobile telephones

• Context
• 6,500 handsets
• 4 million transactions a year

23
Mobile telephones

• Data
• Internal data quality was good but did not
  provide a means of identifying who destination
  numbers belonged to
• Imported 8 million transactions into IDEA
• Got rid of duplicates

24
Mobile telephones

• Focus
• Calls to 190 numbers

25
Mobile telephones

• Identified 190 calls which reduced the
  population size to just under 3,000 transactions
• Used Internet to identify the services provided
  by each 190 number
• Excluded reasonable personal uses such as time
  and weather services
• Identified telephones used to access large
  volumes of adult content / gambling hotlines
• Identified other uses which constituted code of
  conduct issues, for example, competition hotlines

26
Mobile telephones

• Results
• Fraud
• 7 phones calling adult / gambling numbers
• Misuse
• 14 phones calling other services
• Fraud risks
• No current review system
• Observations
• Reporting lost / stolen phones

27
Mobile telephones

• Reporting
• 7 internal fraud investigation referrals
• 14 letters to managers to deal with individual
  minor matters
• Tax Office Executive
• System owners
• Article in weekly staff newsletter

28
Data mining of administered systems

• Longer term projects
• Current project involving BAS processing
• Initial data set 6.5 million transactions
• First cut 3 million
• 11,000 records based on risk algorithm

29
Conclusion

• Focussed on emerging risks
• Flexible approach
• Requires intelligence support
• Work with what you have
• Takes time
• Good planning
• Record keeping
• Information sharing
• Marketing of results

30
Thank You

• Questions?

31
Our commitment to you

• We are committed to providing you with advice and
  information you can rely on.
• We make every effort to ensure that our advice
  and information is correct. If you follow advice
  in this publication and it turns out to be
  incorrect, or it is misleading and you make a
  mistake as a result, we must still apply the law
  correctly. If that means you owe us money, we
  must ask you to pay it. However, we will not
  charge you a penalty or interest if you acted
  reasonably and in good faith.
• If you make an honest mistake when you try to
  follow our advice and you owe us money as a
  result, we will not charge you a penalty.
  However, we will ask you to pay the money, and we
  may also charge you interest.
• If correcting the mistake means we owe you money,
  we will pay it to you. We will also pay you any
  interest you are entitled to.
• You are protected under GST law if you have acted
  on any GST advice in this publication. If you
  have relied on GST advice in this publication and
  that advice later changes, you will not have to
  pay any extra GST for the period up to the date
  of the change. Similarly, you will not have to
  pay any penalty or interest.
• If you feel this publication does not fully cover
  your circumstances, please seek help from the Tax
  Office or a professional adviser.
• The information in this publication is current at
  September 2006. We regularly revise our
  publications to take account of any changes to
  the law, so make sure that you have the latest
  information. If you are unsure, you can check for
  a more recent version on our website at
  www.ato.gov.au or contact us.