Data Sharing for Fraud Detection in Insurance: Challenges and
Possibilities
Carl Christophe Louis Søilen-Knutsen and Bjørnar Tessem
Department of Information Science and Media Studies, University of Bergen, Norway
Keywords:
Data Sharing, Insurance, Fraud Detection, Innovation.
Abstract:
Digital development has opened up new tools to enable innovation, one of the options being data sharing
among businesses. This paper addresses data sharing in the insurance industry and its innovation potential
through a case study from a Norwegian data sharing project. The goal of the studied project is to achieve
cross-company data sharing and with that enabling more efficient insurance fraud detection. We look at what
requirements need to be fulfilled for data sharing to be implemented and what kind of challenges such a data
sharing project meets. We analyse interview data from project participants and systematize their opinions and
impressions regarding possibilities and challenges for data sharing. The case shows that data sharing among
competitors in the insurance industry is hard to realise, very much due to the lack of trust in how the others
will use data, but also due to competition laws and other regulations.
1 INTRODUCTION
Data sharing has great potential in business process
innovation (Richter and Slowinski, 2019). However,
there are in general many obstacles in the way of com-
panies, entrepreneurs and the business community be-
fore they will be able to fully exploit the potential that
data sharing can offer.
The insurance industry is a particular domain
where data sharing may benefit both business and cus-
tomers. A project led by the Finance Innovation busi-
ness cluster in Bergen, Norway (Finance Innovation,
2021), is currently underway adopting data sharing
among insurance companies with the aim of prevent-
ing insurance fraud. The assumption is that a larger
pool of data about customers’ coverage and claims
will improve the ability to uncover fraud. The chal-
lenges are many and the biggest challenges may not
be technical but rather at the legal level. For example,
many of the project participants indicate that Norwe-
gian legislation regarding data sharing is outdated.
Data sharing with personal data is further com-
plicated by the General Data Protection Regulation
(GDPR) within the European Union (EU), which also
applies to Norway. The GDPR makes it pertinent to
question whether the project of Finance innovation
can be implemented at all, because of its restrictions
on personal data. At the same time, the EU recognizes
that data sharing is an important instrument in the de-
velopment of the European economy (Arnaut et al.,
2018).
Several authors have addressed data sharing
among companies, and they all point to the possib-
ilities in various businesses areas (Huttunen et al.,
2019; Celtekligil and Adiguzel, 2019; Tang et al.,
2018). This also includes data sharing for the pur-
pose of insurance fraud detection (Power and Power,
2015). Others are focusing also on the difficulties the
companies may encounter (Eckartz et al., 2014). To
develop the idea of data sharing as a practical innova-
tion tool there has been suggested several frameworks
for organising and enabling data sharing (Grabus and
Greenberg, 2017; Richter and Slowinski, 2019; Eck-
artz et al., 2014). In this research we take a case study
approach, and address the issues of data sharing met
in Finance Innovation’s data sharing project. We use
the interview data we have gathered to further develop
our understanding of the data sharing concept, and
aim to develop a richer understanding of the problems
and opportunities related to data sharing.
We start with a background of current research
on data sharing, and continue with a description of
the Finance Innovation project ”Fra data til innsikt”
(From data to insight). We go through our data col-
lection approach, and try to draw some new insights
in a discussion section, before we conclude.
Søilen-Knutsen, C. and Tessem, B.
Data Sharing for Fraud Detection in Insurance: Challenges and Possibilities.
DOI: 10.5220/0010982300003179
In Proceedings of the 24th International Conference on Enterprise Information Systems (ICEIS 2022) - Volume 1, pages 93-99
ISBN: 978-989-758-569-2; ISSN: 2184-4992
Copyright
c
2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
93
2 DATA SHARING
Arnaut et al. (Arnaut et al., 2018) describe, in
their report to the European commission, data shar-
ing as ”The process by which a company re-uses data
from another company, which is not a direct market
competitor, for its own business purposes (excluding
contractor-subcontractor relationships). These data
were either accessed for free or acquired against some
remuneration or other kind of compensation, includ-
ing the provision of a service.” The EU report sug-
gests that such data sharing may contribute signific-
antly value to companies and society, in terms of new
business models, new products, and improved internal
efficiency in companies. Within a government open
data perspective, it has been shown that access to
more data gives a small, but clearly significant in-
novation value (Jetzek et al., 2013). Even though the
data sharing concept we are referring to here is not
100% the same thing, authors repeat the message also
for business-to-business (B2B) data sharing; there are
huge gains to be found (Huttunen et al., 2019).
Power and Power state that insurers around the
world can save large sums if data can be shared in-
ternally in the industry. To achieve this, it is neces-
sary that the process be streamlined so that sharing
can take place as efficiently as possible and fraud can
be uncovered more easily (Power and Power, 2015).
They suggest that one could use data ranging from in-
dividual insurance claims, overall frequency of insur-
ance use, types of insurance activities, individual be-
havioral data, and individual social media activities.
Then probability based analyses of the data should
enable targeting of the suspicious cases. There has
also been parallel reflections about the possibilities
and challenges in health care, mentioning for example
the potential in sharing data about rare cases (Tang
et al., 2018).
Huttunen et al (Huttunen et al., 2019) did an em-
pirical study documenting B2B data sharing in Fin-
land. 49% of companies in Finland did data sharing,
and they had significant business gains from this; in
particular from optimised operations. However, the
need for data market places was not so accentuated,
as data sharing often was established through bilat-
eral agreements. Celtekligil and Adiguzel (Celtekli-
gil and Adiguzel, 2019) has also verified innovation
value from data and information sharing in tech in-
dustry, in particular if management support is high.
A few frameworks for data sharing addressing dif-
ferent aspect have been suggested. Richter and Slow-
insky emphasise the need for data sharing platforms,
and suggest to develop a framework for exchange of
data similar to the FRAND model for exchange of
patents (Richter and Slowinski, 2019). Eckartz et al.
(Eckartz et al., 2014) is more oriented to the prac-
tical issues of data sharing and have developed a four
step framework for the process of sharing data, in
particular addressing constraints. These constraints
are Ownership, Privacy, Economic, Data Quality, and
Technical. They used three cases from logistics to il-
lustrate their model. Grabus and Greenberg (Grabus
and Greenberg, 2017) have collected a number of
data sharing agreements, identified commonalities,
and developed a framework that could be used to
automatically generate agreements taking care of pri-
vacy, legal concerns and other restrictions.
3 FROM DATA TO INSIGHT
Finance Innovation
1
(FI) is an initiative from the fin-
ancial industry in Bergen, Norway to start a centre for
innovation in Fintech. FI was started in early 2018
and one of their initial actions was to develop a col-
lection of projects. The insurance industry signalled
an interest in sharing data, motivated by an assump-
tion that more data would give more precise mod-
els for fraud detection. This was especially true of
the companies engaged in asset insurance, e.g., the
insurance company Tryg
2
, which already had fraud
models, but wanted more data. But also within the
fields of people insurance (life, health, pension) there
was interest. Yearly there is uncovered fraud claims
for more than 300 million Norwegian Kroner in Nor-
way, and the insurance business actors think the un-
discovered fraud has a value many times higher than
that (Finans Norge, 2019).
Fraud models are algorithms that insurers use to
assess whether an insurance claim is legitimate or is
an attempted fraud. Such models are used by all mod-
ern asset insurance companies, and they use machine
learning and data to improve their own fraud detec-
tion. The assumption is that the more data a fraud
model has access to, the better the model will be
able to detect an insurance fraud. At the same time,
people insurers such as DNB Liv
3
want to develop
better computer-based models in life, health and pen-
sion insurance. Today, investigations of fraud suspi-
cions against insurance claims in people insurance are
something that is initiated based on assessments from
case officers only.
From this the project ”Fra data til innsikt” (From
data to insight) was started with the goal of facilit-
ating the sharing of claims data and fraud data from
1
https://financeinnovation.no/
2
https://www.tryg.no
3
https://www.dnb.no/forsikring/personforsikring
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
94
the various insurance companies (Finance Innovation,
2021). The project started with participants from the
finance industry, including Tryg and DNB Liv, con-
sulting companies such as Webstep and PwC, and
researchers from the University of Bergen, the Nor-
wegian School of Economics and Business Admin-
istration and Western Norway University of Applied
Sciences. Engagement has been high at times, but at
present the activity is low. For the time being, there is
work going on with a so called minimal viable proto-
type (MVP) within car insurance, involving only two
companies.
Within the project, it immediately became clear
that there were challenges on several fronts. There
were business and legal challenges, technological
challenges, challenges with what data can be shared,
and challenges with analytical methods. Over the past
two years the project worked on the various issues,
and received strong support from the subdivision for
insurance in Finans Norge
4
(collaboration organisa-
tion for the Norwegian finance industry), and inside
the participating finance industry companies.
However, in the last year and longer, the project
has had little progress. Thus, the project can hardly be
said to be a success. But perhaps that particular fact
makes it more interesting to uncover what factors con-
tributed to today’s status of the project. Hopefully a
case study of the project can provide important learn-
ing and understanding to aid future innovation pro-
jects with data sharing.
Thus, through a case study (Benbasat et al., 1987)
of this project we may provide some valuable answers
to these questions: What challenges in data sharing
for fraud detection does the insurance industry have
to address? What solutions may contribute to realize
the innovation potential of data sharing for fraud de-
tection?
4 DATA REGARDING THE
PROJECT
We have gathered only qualitative data about the pro-
ject. The most important data are interviews with pro-
ject participants, but there is also some documentation
in the shape of internal project emails.
The emails used in the analysis are provided from
one of the authors who had an active role in the pro-
ject. These emails also had attached some files (texts,
presentations) relevant to the project. These emails
and the attachments were used to get an overview of
the earlier stages of the project and the direction the
4
https://www.finansnorge.no/en/
project took.
The main data for our analysis are semi-structured
interviews with project participants. The participants
have been selected on the basis of the second author’s
knowledge of the project and on suggestions from
Finance Innovation’s project managers. It was a goal
to have a diverse professional background in the re-
spondent group. All the requested respondents were
positive to participating in the interviews. The inter-
view guide is found in the Appendix.
We interviewed seven respondents. Of these, six
are still active in the project, while one has been
out of the project for a couple of years. Almost all
interviews were conducted digitally using Microsoft
Teams due to the risk of coronavirus. One inter-
view was conducted over the phone. Participants were
made aware that they were recorded on audio or video
recordings.
The respondents were
1. Project manager, Finance Innovation
2. Manager, tech consultancy company
3. Manager, business intelligence, insurance com-
pany
4. Data scientist, insurance company
5. CEO, Finance Innovation
6. Lawyer, consultancy company
7. Project manager (previous), Finance Innovation
5 ANALYSIS OF INTERVIEWS
AND EMAILS
In the following the findings from the interviews are
organised into seven topics:
• misaligned expectations
• GDPR and other regulations
• public support
• technical solutions
• facilitating the project
• the future of fraud
• other opportunities
During analysis, statements were coded with terms
matching topics in the interview guide, but we also
coded for other issues. Topics that emerged as im-
portant during the analysis were misaligned expecta-
tions, technical solutions, and facilitating the project.
The other four topics came out of issues addressed
directly in the interview guide. During interviews the
questions about GDPR, public support, fraud in the
Data Sharing for Fraud Detection in Insurance: Challenges and Possibilities
95
future, and new business opportunities got the richest
answers.
5.1 Misaligned Expectations
In the project’s earlier phases, it was difficult to get
the actors to agree on the scope of the project. The
project initiative came from Finance Innovation, and
to a less extent from the participants. The earlier
meetings was meant to create some consensus, and
data sharing was from the start the main theme, and
then insurance fraud was seen as one possible arena
for data sharing.
There was a lot of unstructured discussion about
how things could be done and it was difficult to reach
agreements, and the emails show that the same topics
are recurring from meeting to meeting. Respondent 2
said, ”there was a working group discussing and there
somehow never became anything real out of it”. The
various participant companies sent people with dif-
ferent background. For example, one insurance com-
pany sent a data analyst, whereas another sent a busi-
ness responsible. One aspect was the varying size of
the companies. Some are market leaders, and others
are smaller niche companies. One respondent pointed
to the lack of funding to the project from the parti-
cipating companies: ”We need money, we need fin-
ancial commitment from the actors, that’s what it’s
really about”.
5.2 GDPR and Other Regulations
All respondents considered GDPR to be a challenge
to the project’s feasibility. However, it was not con-
sidered the biggest challenge among all the respond-
ents. But no one wanted to push the regulatory limit-
ations, and this meant that important parts of the data
would need to be excluded from data sharing. Re-
spondent 1 said: ”We have sacrificed quite a bit of
explanatory power in order to stay on the right side
of the law”. They also mentioned other issues, like
competition regulations in Norway, which state that
it is not legal for businesses to cooperate if they have
more than 30% market share. There are also laws that
restrict the sharing of information among finance in-
dustry companies.
It was stated that both the GDPR and the Nor-
wegian competition laws are unclear and that this
presents significant challenges for the project. A sim-
ilar project has not been done earlier in Norway. This
means that the interpretation of laws lacks precedents,
further complicating the understanding of the laws
and consequently the progress of the project.
5.3 Public Support
During the interview the government’s role in the pro-
ject was discussed, and several of the respondents
mentioned Datatilsynet
5
(Norwegian Data Protection
Authority). They have set up a regulatory sandbox
where businesses can try out solutions without being
afraid of breaking regulations. This project would be
a good case for the sandbox, and would ensure that
all the issues with regulations could be put aside for a
while. Respondent 3 said, ”the regulatory sandboxes
are a place where you can test out a project, without
fear of reprisals, you might say. Where you work
in cooperation with the regulatory authorities.” One
of the respondents mentioned that this project has so
high societal value that it should allow for exemptions
from regulations.
5.4 Technical Solutions
Recently the project has decided to focus on creating
a MVP solution for car insurance. Only two of the
initial insurance companies are involved in this (at
the last time we checked), although one would have
wanted more to make a solution viable. The envi-
sioned, so called black box, solution is one where the
companies will be able to see their own data only. But
they may learn fraud models using all the available
data. It is assumed that this will solve the regulat-
ory challenges. However, this solution is according
to respondent 6 semi-optimal. He said: ”The solu-
tion, considering legal issues, would be to anonymise
the data sets. Then you can do whatever you want
with data and you can keep them for how long you
want. But they were not able to do that. So I think the
black box solution was a very good way to be able to
actually give that AI enough data to train a model”.
Respondent 4 was more positive, and focused on how
the MVP reduced chances of data misuse, but with
a realistic twist: ”It is quite difficult to completely re-
move the ability to misuse the data and we have talked
about different solutions to that.”
5.5 Facilitating the Project
As mentioned the project was initiated from FI, as a
response to a brain storming process among finance
industry actors. And having a facilitator like FI was
essential for the project to get going. The importance
of Finance Innovation being a non-profit organization
in the middle was pointed out as a trust builder among
participants. For example, using the premises of FI
5
www.datatilsynet.no
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
96
allowed companies to meet on neutral ground and co-
operate. It would have been different if such a meet-
ing were to take place at one of the companies. In
particular, trust was seen as important because there
are fears that the data may be misused by dishonest
forces. One of the respondents said that the weekly
meetings and were an important contributor to creat-
ing trust.
But still the project progressed slowly. Respond-
ent 3 suggested this was due to the lack of an early
”minimum solution, good enough to demonstrate to
the managements of the various insurance compan-
ies that this is something that will bring great value.
Which will be worth investing in.” The minimal solu-
tion was hard to do because of lack of example data
from the various companies. In the end, they seemed
to not dare to share data for an experimental solution.
The companies had a strong scepticism towards shar-
ing data because of a fear that competitors might get
insights into their business models. It also was clear
that it was difficult for them to set aside the people re-
sources to prepare data. This again relates to the fin-
ancial commitment from the actors that we also men-
tioned previously in the analysis.
5.6 The Future of Fraud
The respondents were invited to reflect on what the
effect of such data sharing could be. Respondent 3
said ”There is quite a few hundred million a year
in costs so there is clearly a potential in preventing
fraud. Thus in the utmost consequence you give a
cheaper insurance to the honest customers”. But there
is no indication that such sharing of data will actu-
ally lead to more uncovered fraud. This remains to be
seen. Still, some speculated about the possibility from
data sharing to uncover new and more advanced fraud
attempts. They were sceptical about the customers’
ability to change their behavior as more fraud would
be uncovered. The arguments was that they do not
have insight into the algorithms, but also that insur-
ance fraud is mostly not done by professional crimin-
als.
5.7 Other Opportunities
Some respondents envisioned that data sharing also
could give other opportunities than improved fraud
detection. Respondent 2 reflected on the possibility
to compete with international actors who have a large
advantage in amounts of data, like Google and Face-
book: ”this is an opportunity for the Norwegian insur-
ance industry to meet competition with any interna-
tional players who come with a large data advantage”.
Although related to data sharing only indirectly, auto-
matizing decision processes was mention by several
respondents. But they did not see any other business
opportunities or new products.
6 A NOT SO SUCCESSFUL DATA
SHARING PROJECT
The ”From Data to Insight” project was started with a
diverse group of participants, with the aim of sharing
data among companies in insurance. The project de-
cided to focus on sharing data for the detection of in-
surance fraud, based on assumptions about a potential
similar to what Power and Power argued for (Power
and Power, 2015). That is, more and richer high qual-
ity data shared among competitors would enable auto-
mated fraud detection, and significant reductions of
losses. The project was not successful compared to
the initial expectations, and has ended up with only
two partners sharing data, aiming towards a somewhat
restricted black box solution.
If we look at EU’s definition of data sharing (Ar-
naut et al., 2018), we notice immediately that this
does not include data sharing among competitors,
which the FI project actually aims to do. And perhaps
this is also the reason why the project never really
took off. Eckartz et al.’s (Eckartz et al., 2014) list
of constraints had Ownership on the top, and most
likely there was some lack of trust (or not enough
trust) among the companies regarding how others
would use the data to infer business models. The com-
panies actually saw a big problem in sharing these
data. This, together with the regulations imposed
by Norwegian competition laws put in effect due to
being competitors, actually created some Ownership
hurdles in the start that the project never managed to
pass. Privacy was also considered an important is-
sue, but since actually no data was really shared in
the sense that competitors could see each other data,
it was less of a problem than the competing company
issues.
Another aspect that hindered the development of
the project, was the lack of technical solutions show-
ing its viability. This most likely caused less support-
ing management in the participating companies. They
wanted to see how value could be realised before
committing to the data sharing idea. An early regulat-
ory sandbox solution demonstrating the value would
probably have helped in engaging management more.
Celteklicil and Adiguzel (Celtekligil and Adiguzel,
2019) have shown that, as in most information sys-
tem projects, management support and involvement is
critical to success. The insurance business is conser-
Data Sharing for Fraud Detection in Insurance: Challenges and Possibilities
97
vative, and wants to see clear value from initiatives.
Huttunen et al. (Huttunen et al., 2019) showed
that shared spaces for data sharing are not necessar-
ily needed, bilateral solutions may give just as high
value. The advantage is that those may result in less
complex agreements, and remove some of the more
challenging competition and privacy issues. Eventu-
ally this was the outcome of this project, two compan-
ies collaborating on data about car insurance.
In the initial research questions for the project we
aimed to get insights into what kind of solutions were
possible in such a project. What we observed is that
beyond the restricted black box solution, which is
not yet operative, there are basically are no solutions
found yet. Very much stopped due to the inability to
handle the initial constraint of ownership. Perhaps, if
one had a focused on a demonstration solution from
early on, there would been more engagement in trying
to solve the ownership issue.
This project is a case study with a limited set of
interview respondents, and will of course not general-
ise to other situations, and in particular to other pro-
jects on data sharing for insurance fraud detection.
However, it illustrate some of the problems such a
project may meet, and emphasise the need to focus
on early demonstration of a viable system, focus on
handling trust among data sharers, and commitment
among managers to assign the necessary people re-
sources. As mentioned, insurance fraud has large so-
cietal costs, and solutions to reduce it will have great
value for companies, customers, and society in gen-
eral.
7 CONCLUSION
This paper describes a case study of a data sharing
project in the insurance industry, where data sharing
was hoped to be a driver for innovation. The pro-
ject was in large parts unsuccessful. We have ob-
served how data sharing among competitors becomes
problematic due to several forms of regulations, di-
vergent interests, lack of trust, lack of management
support, and other issues. The case study shows that
data sharing among companies is difficult to realise,
even though there is a large potential for economic
gains for all parties. One observation is that early
demonstrators of solutions, for example, in a regulat-
ory sandbox, could contribute to increased trust, man-
agement support, and engagement.
REFERENCES
Arnaut, C., Pont, M., Scaria, E., Berghmans, A., and Le-
conte, S. (2018). Study on data sharing between com-
panies in Europe. everis Benelux.
Benbasat, I., Goldstein, D. K., and Mead, M. (1987). The
Case Research Strategy in Studies of Information Sys-
tems. MIS Quarterly, 11(3):369–386.
Celtekligil, K. and Adiguzel, Z. (2019). Evaluation of Data
Sharing in Production Firms and Innovation Orienta-
tion in The Effect of Management Capability on Op-
erational Performance. Procedia Computer Science,
158:781–789.
Eckartz, S. M., Hofman, W. J., and Van Veenstra, A. F.
(2014). A Decision Model for Data Sharing. In
Janssen, M., Scholl, H. J., Wimmer, M. A., and Ban-
nister, F., editors, Electronic Government, Lecture
Notes in Computer Science, pages 253–264, Berlin,
Heidelberg. Springer.
Finance Innovation (2021). Detection of Insurance Fraud.
https://financeinnovation.no/innovation-projects/
detection-of-insurance-fraud. Accessed: 2021-08-17.
Finans Norge (2019). Forsikringssvindel i Norge - svikstat-
isikk 2019 (Eng: Insurance fraud in Norway - fraud
statistics 2019).
Grabus, S. and Greenberg, J. (2017). Toward a Metadata
Framework for Sharing Sensitive and Closed Data:
An Analysis of Data Sharing Agreement Attributes.
In Garoufallou, E., Virkus, S., Siatri, R., and Kout-
somiha, D., editors, Metadata and Semantic Research,
Communications in Computer and Information Sci-
ence, pages 300–311, Cham. Springer International
Publishing.
Huttunen, H., Sepp
¨
al
¨
a, T., L
¨
ahteenm
¨
aki, I., and Mattila, J.
(2019). What Are the Benefits of Data Sharing? Unit-
ing Supply Chain and Platform Economy Perspect-
ives. Technical Report 93.
Jetzek, T., Avital, M., and Bjørn-Andersen, N. (2013). Gen-
erating Value from Open Government Data. ICIS 2013
Proceedings.
Power, D. and Power, M. (2015). Sharing and Analyzing
Data to Reduce Insurance Fraud. MWAIS 2015 Pro-
ceedings.
Richter, H. and Slowinski, P. R. (2019). The Data Sharing
Economy: On the Emergence of New Intermediaries.
IIC - International Review of Intellectual Property and
Competition Law, 50(1):4–29.
Tang, C., Plasek, J. M., and Bates, D. W. (2018). Rethinking
Data Sharing at the Dawn of a Health Data Economy:
A Viewpoint. Journal of Medical Internet Research,
20(11):e11519.
ICEIS 2022 - 24th International Conference on Enterprise Information Systems
98
APPENDIX
Interview Guide
• Tell me about your professional background and
how you got into the finance innovation project?
What kind of tasks do you have on your project?
• What do you see as the biggest challenges facing
the project today?
– Do you think of GDPR as a challenge to your
project? Why? Why not?
– What challenges does the GDPR present?
– Do you think that Norwegian regulations
present challenges for the project? In case,
what challenges?
– Do market participants pose any challenges? In
case, what challenges?
• What do you think it takes for the project to be
realized?
– Do you think trust is an important factor?
– Why is trust important? How can trust be im-
proved?
– Can the insurance industry learn anything from
the banking industry in terms of data sharing?
The have managed to cooperate on systems
such as BankID. What do you think you can
learn from it? What can be transferred from the
banking industry to the insurance industry?
– Do you think this is something the Norwegian
authorities can do to realize the project? In case
what?
• What opportunities do you think the project can
give the insurance industry?
– Do you think that the insurance industry will be
more efficient with data sharing?
– Do you think it can give the insurance industry
new business opportunities? In case what?
• How do you think the project could help uncover
insurance fraud?
– To what extent do you think it could help un-
cover insurance fraud?
– How will it support the process of solving in-
surance fraud?
– Are there any special types of fraud that are go-
ing to be more easily uncovered than others?
– Could we see new forms of insurance fraud if
the project is successful? In case what?
Data Sharing for Fraud Detection in Insurance: Challenges and Possibilities
99
