ELIMINATING TELCO FRAUD WITH SELF LEARNING MACHINES
ELIMINATING TELCO FRAUD WITHSELF LEARNING MACHINES
READ THIS WHITEPAPER TO UNDERSTAND:
• What is required to reduce fraud levels in a Telecom company• HowWiseAthenadeliverseffectiveTelecomfraudprotectionusingmachinelearningandcognitivevisualization• HowtheSaaSmodelisstraightforwardtointegrate,requiresnoendusermaintenanceanddelivers a ROI measurable in days
”They made me see in ten minutes whatIhadnotrealizedaboutmyfraudin twelve years.”FRAUD MANAGER, FORTUNE 500 COMPANY
Introduction - Telecom fraud: a persistent challenge 4
Fraud use case: SIMbox 6
Self learning machines for detecting fraud in near real time:
a perfect combination 6
Fraud in the Telecom industry: is it really that bad? 8
The inner workings of SIMbox Fraud 10
The need 12
How Wise Athena improves your fraud detection 13
What is missing in existing solutions? 14Fundamentals of our solution 15
How cognitive analytics eliminates SIMbox fraud 15
Finding the sources of fraud 17
Machine learning results for your Telco 19
How smart visualizations improve your business decisions 22
Business Results in Four Weeks 24How it works for you 25
It only gets better with time 26
Your savings with Wise Athena 27
The next step for your fraud 28
Table of Contents
www.wiseathena.com | [email protected] 3
www.wiseathena.com | [email protected] 4
Introduction
Telecom fraud: a persistent challenge
wise athena
Telecom industry efforts to address the growing problem of fraud
fall short. Successful fraud detection requires huge data-processing
and fast reaction time. These are challenges in themselves. However
many companies are still not aware of the new –and more effective-
technologies that are now available.
Wise Athena applies cognitive analytics to bridge the gap between
current fraud solutions and the most effective technology available.
We make Machine Learning and smart visualizations work together
to deliver an 800% increase in human perception. We answer
questions as yet unimagined. Our results are actionable and visual.
Very complex data can be understood at a glance.
A Wise Athena’s smart visualization of a telecom cellular network
IMEI
Called SIM
Calling SIM
www.wiseathena.com
Although Telecom fraud can take myriad forms, we will examine one
variety that remains a significant challenge: SIMbox.
SIMbox fraud is considered one of the most serious types of fraud for
telecommunications operators globally. Consequences range from
billions in financial losses to the straining of cellular networks caused
by increased fraudulent traffic.
Despite having allocated significant resources to prevention,
conventional rule-based SIMbox fraud prevention has consistently led
to incorrect outcomes. The industry needs a new solution.
• Traditional detection methods identify leaks in revenue, but
make it hard to accurately pinpoint the source.
• False positives can lead to the operator taking action on
legitimate users.
• Fraudsters break even at twenty three minutes of fraudulent
calls. They understand traditional detection methods and change their
patterns accordingly.
Wise Athena uses machine learning, cognitive analytics and smart
visualizations to bring a new solution to the Telco sector. Our
technology maps the essential behavior of network data to identify
anomalies. It brings celerity and accuracy to fraud detection.
Our cognitive system learns by understanding relationships in large
volumes of data points. This ever-learning capacity allows our data-
processing infrastructure to automatically keep itself up-to-date. It
learns the fraudsters’ new ways to conceal fraud.
This combination of cognitive technology and smart visualizations
Fraud use case: SIMbox
Self learning machines for detecting fraud in near real time: a perfect combination
www.wiseathena.com | [email protected] 6
produces actionable graphs. Hidden patterns are revealed, making
data exploration a breeze. No longer is a team of data analysts
required to know what is happening in your network. You will actually
be able to see the shape of fraud.
At Wise Athena we have harnessed the inherent power of
visualizations. Because we tailor them to your data, our visualizations
will increase your perception eight times over traditional methods. It
becomes easy to action clear, complete, and correct results.
And what are the results?
SIMbox fraud is neutralized and no longer profitable for illicit
operators. With as little as seventeen minutes of fraudulent calls, false
positives are reduced with an accuracy 10,000 times greater than the
traditional methods (from 1% to 0.001%.) Our 24x7 platform returns
results in one minute, so SIMbox fraud can be seen happening in near
real time.
Because this is provided as a service (SaaS), you benefit from our
responsiveness. Economies of scale and agility provide results at a
tenth of the cost, and a sixth of the time compared to other providers.
www.wiseathena.com | [email protected] 7
www.wiseathena.com | [email protected] 8
Fraud in the Telecom industry can cause great damage to an
organization’s image and operations. The different illicit activities
targeting operators can lead to significant increases in costs, worsened
brand image and customer dissatisfaction.
In the Telco industry, the lack of effective fraud detection tools has led
in many cases to the loss of control over fraudulent behaviour, allowing
the fraudsters to create a sustainable business model at the companies’
expense.
SIMbox fraud (Subscriber Identity Module box), also known as bypass
fraud or voice traffic termination fraud, rates among the top five fraud
types globally in the Telco industry. That is both in impact and speed
of growth1. Illicit use of cellular networks is one of the most lucrative
varieties, and it is very costly to mobile operators.
These schemes appropriate international voice calls and reroute them
over the Internet to a box with multiple prepaid SIM cards. This SIMbox
subsequently transforms the incoming signal back into a mobile call
that is charged at a local rate on termination. Legitimate operators
Fraud in the Telecom industry: is it really that bad?
wise athena
1. CFCA 2013 Global Fraud Loss Survey - http://www.cfca.org/fraudlosssurvey/
involved in the connection are then unable to apply charges for the
international routing.
Every year, operators incur in losses calculated in $40Bn worldwide2
as a consequence of this type of crime. Moreover, networks are put
under increased strain due to the high level of traffic that targeted
nodes support. Ultimately, quality and reliability are compromised.
Use of a SIMbox to bypass international gateways is categorised as a
crime by Telecommunications Authorities in many countries. These
activities remain not only a serious problem for telecommunications
operators but also for governments and authorities:
• Providers of this bypass connectivity are usually not licensed.
• They can be a threat to national security, given the lack of
control of operators and authorities over their actions.
• Substantial losses on tax revenue.
For all these reasons, detection and prevention of SIMbox fraud
is now a priority for operators. However, conventional solutions
consistently fail at accurately locating and isolating the occurrences of
this phenomenon.
The solution presented in this paper helps eliminate SIMbox
fraud by applying machine learning, cognitive analytics and smart
visualizations. Our experiments using data from a major operator in
a market with a high prevalence of SIMbox have shown that these
technologies succeed. They correctly identify and isolate fraud
occurrences. Moreover, our cognitive algorithm systematically learns
from every outcome. The system becomes more effective with every
interaction.
Wise Athena uses accurate, efficient and evolving technologies to
identify fraudulent calls that bypass your gateways and cause lost
revenue.
2. Fraud Survey Report 2010, KPMG, 2010
www.wiseathena.com | [email protected] 9
www.wiseathena.com | [email protected] 10
SIMbox fraud takes advantage of the difference in price between local
and international calls. Fraudsters use the Internet to send one call from
one country to another thereby avoiding the legal international inter-
connection gateway charges. This bypass normally occurs by routing the
calls via VoIP to a SIM in the country where the receiver resides.
The fee callers pay is higher than that for a local connection; yet conside-
rably lower than the legitimate one. This difference is a source of reve-
nue to the fraudster.
Legitimate network operators suffer sizeable losses of revenue as a
consequence.
The inner workings of SIMbox Fraud
wise athena
In the figure above, there are two possible paths when a call is
initiated. In the legitimate scenario, the call is routed through the
operator’s network in the country of origin. Then it is sent to the
network of the country of destination where a second operator
delivers the call to the receiver. On termination, the relevant charges
are made and revenue is distributed among the operators accordingly.
In the fraudulent scenario, once the call is made, the illegitimate
operator misappropriates the signal. They then send it to a SIMbox
in the country of destination using the Voice over Internet Protocol
(VoIP). The SIMbox injects the signal back into the cellular network
and finally it reaches the receiver’s end in the form of a local call
charged at a significantly lower rate.
The setback for the operator is not only financial. The fraudulent
service usually carries a substantial reduction in the quality of the
connection resulting in a call of substandard quality. In addition, the
rerouting implies a lack of transparency for the users e.g., the calling
number on the receiver´s display does not correspond to the real user
originating the call.
SIMbox fraud is more prevalent in countries where international
calls are considerably more expensive than local calls, e.g. where
the gateways are monopolised by the government. It is especially
common in Africa, Middle East, Latin America and parts of Asia.
www.wiseathena.com | [email protected] 11
www.wiseathena.com | [email protected] 12
The Telco industry needs to move on from a reactive model to a
predictive model that not only detects existing fraud but also creates
a deterrent for future. It must completely neutralize the fraudster’s
business case.
Traditional approaches to SIMbox fraud detection have showed limited
effectiveness. This is due to their inability to prevent future fraud.
Existing methodologies (e.g. Fraudview) involve the manual update of
SQL queries by a team of analysts. This occurs every time new instances
of fraud are discovered. Updates are effective at neutralizing existing
fraud but the system as a whole has no capacity to prevent future
varying occurrences.
The shortcomings of this approach are many; a team of experienced
analysts available 24/7 represents a big expense for operators. In
addition, the reactive nature of the methodology means that fraudsters
are always one step ahead. They evolve new ways to evade control.
There is a need to completely neutralize the business case of the
fraudster; reactive methodology does not fulfill it.
The Needwise athena
Our system uses machine learning, cognitive analysis and smart visualizations to identify variables that will
accurately pinpoint the SIMbox fraud you are suffering.
These technologies allow us to deliver a superior solution that makes big improvements in performance,
maintenance and scalability. Ease of implementation means that you can start enjoying the benefits without
delay.
YOU DON’T PROGRAM
IT, IT LEARNS
Machine learning
eliminates manual query
updates and constant re-
programming. With only
a small amount of data
to train, the system then
starts producing results.
DATA EXPLORATIONWise Athena uses smart visualization technology for producing a graphical representation of your data. These powerful images stimulate data discovery and reveal properties in your data that remain hidden for traditional data visualization tools (static pie charts, spreadsheets.)
EASE OF SETUP
Wise Athena’s suite
of analytics tools run
on external dedicated
servers and do not
require installation by the
operator. Money and time
are saved.
SPECIFIC FOR TELCOS
Wise Athena’s tools are
specifically designed for
Telcos while conventional
Telecom fraud prevention
tools have generally been
part of generic packages.
NO LEARNING CURVE
The system is accessible
to multiple user
types and skill levels.
Actionable insights
extracted from data are
easy to understand and
implement. A dedicated
team of data analysts is
not required.
TRANSPARENT PRICING
Wise Athena uses a
transparent and results-
based pricing system.
The operator is charged
proportionally to the
results achieved. This
avoids the hefty license
fees of traditional
solutions.
NO NEED FOR OWN
INFRASTRUCTUREWise Athena uses the SaaS (Software as a Service) model. This means that Wise Athena takes care of time-consuming and expensive hardware and software maintenance costs.
THE SYSTEM ADAPTS
TO YOU
Wise Athena adapts
to the structure of the
operator’s data and not
vice versa. The system is
ready to start processing
data as soon as it is made
available.
How Wise Athena improves your fraud detectionwise athena
Advantages of Wise Athena’s solution:
www.wiseathena.com | [email protected] 14
Traditional detection of SIMbox fraud requires monitoring calling
patterns and using test calls to find a leak in revenue due to bypass.
These efforts are effective for acknowledging the issue, but insufficient
for identifying and addressing it.
The volume of services and technologies available to fraudsters makes it
a constantly evolving and multi-faceted issue. Actions taken to eliminate
fraud modify the fraudsters’ behaviour, leading to ever-changing
patterns and renewed threats to operator revenue.
In some markets existing solutions are so ineffective and costly, it is
easier to accept the losses than resolve the problem.
Another issue that compromises the operators’ ability to effectively
addressing fraud is the serious commercial implications of a false
positive. A legitimate customer may see their line cancelled for no
apparent reason. This compromises the reputation and reliability of the
operator. Accuracy is critical.
The challenge is predicting what new fraud patterns will emerge , then
quickly acting to stop them before they become profitable. Telecom
companies face major difficulties when confronted by the exponential
growth of data. Fraud detection can become an immense task.
What is missing in existing solutions
wise athena
Wise Athena uses advanced analytics supported by quantitative
methods to produce insights that traditional approaches to fraud
detection are unlikely to discover.
No matter the fraudsters’ constantly evolving patterns, there are
some elements that escape their control:
• Their past behavior
• Legitimate SIMs behavior
• Duration of fraudulent calls
• Time of the calls
• Location of the base stations
Wise Athena uses these fraud-defining features as classifiers to
detect fraudulent use of the network. Legitimate users are filtered
out, even if their patterns are comparable to those of the fraudster.
Illegitimate users are also limited by a lack of control over other
elements, such as the cost and structure of their business model.
Actions taken for evading fraud-detection measures bring added cost
to their operations and greatly reduce their ability to remain hidden.
Wise Athena’s system integrates all this knowledge into a SIMbox
fraud-detection tool that is accurate, efficient and self-improving.
A system that identifies current fraud and its defining patterns places
a big hurdle for the fraudsters’ business model. They must step up
efforts to evade control and see how changes in the pattern are
discovered before reaching breakeven point. Ultimately, the goal of
making fraud unprofitable is reached.
Use case data has been trained on to a Tier 3 operator operating in a market where SIMbox fraud was unresolved.
Wise Athena employed five hundred decision trees to characterize
the SIMs with high probability of being used for SIMbox fraud.
Fundamentals of our solution
How cognitive analytics eliminates SIMbox fraud
www.wiseathena.com | [email protected] 15
www.wiseathena.com | [email protected] 16
Looking at multiple dimensions we identified markedly different
behavior in non- fraudulent SIMs vs. SIMs suspected to be used in a
SIMbox. Here are a few examples:
These graphs show the same data with different dimensions and
perspectives. They are analyzed until a clear picture of the fraud
occurrences emerge.
In this case, the operator sent its network data for processing to
Wise Athena’s engine. Our algorithms mapped the structure of the
network and then represented fraud occurrences with red circles
and legitimate users with black circles. The observation of data
from multiple different perspectives and using multiple dimensions
ensures that fraud is clearly identified after a comprehensive refining
of results. This keeps false positives to a minimum while unearthing
more SIMbox than any other tool on the market.
Each graph represents a different dimension. For example: percentage
increase in calls during weekends, calls between 23:00 and 23:59, call
activity during weekdays compared to weekends, etc. All dimensions
are abstractions of customer behaviour that allow for a cumulative
refinement of results. By putting the most significant ones together,
we gain a clearer understanding of the validity of the results.
www.wiseathena.com
The best way of identifying fraudulent users is by comparing it with
legitimate behavior of normal customers.
Human interactions can be traced very accurately using mobile call
data.
Consider a simple network of people as depicted below, with a mix of
personal and professional relationships.
From the operator’s point of view – i.e. that of its data – the picture
is a bit more complex. E.g.: users make calls from different locations,
sometimes using different numbers – from the work phone or from
the personal one – perhaps using different devices, etc.
Finding the sources of fraud
www.wiseathena.com | [email protected] 17
We could consider one more level of human interactions: Illegitimate users behave differently; they are physically located in
one static point and connect to very few base stations. They also have
a disproportionally high rate of initiated calls to received calls.
This conceptualization is easy to understand at a small scale.
However, the depiction of these relations from a technical point of
view would look very complex for a human mind.
Making a decision based on interpreting this information would be
even harder.
www.wiseathena.com | [email protected] 18
We can identify fraud with a high level of accuracy and provide the
tools to make this information actionable using smart visualizations.
Wise Athena depicts the information in a graphical manner that:
• Conveys large amounts of information,
• Is accurate and intuitive to the reader, and
• Supports decision-making.
Our Machine Learning algorithms combined with smart visualizations
identify relations among data points and provide results such as these
shown below.
These results are for actual Telco companies3.
Machine learning results for your Telco
3. Results from actual Telco company. Name cannot be disclosed due to confidentia-
lity agreements. References can be provided upon customers’ request.
In this image each green dot represents a number called from one SIM,
and each orange dot at the end of each orange line represents the IMEI
(device identifier) for that SIM.
These globules are easily identified as non-fraudulent behavior.
www.wiseathena.com
www.wiseathena.com | [email protected] 19
www.wiseathena.com
In the case of SIMbox fraud, the visualization of the results yields a
remarkably different image.
This is an example of graph which identifies abnormal activity at a
glance.
We can clearly see a pattern that differs from non-fraudulent behavior.
IMEI
Called SIM
Calling SIM
www.wiseathena.com
In these smart visualizations the size of the green dots is proportional
to the volume of traffic on that SIM; likewise, the size of the orange
dots is proportional to the number of SIMs using an IMEI. The image
depicts orange dots in non-fraudulent behavior so small that they
cannot be even seen as such – i.e. non-fraudulent IMEIs do not have
heavy traffic.
The abnormal area, however, shows some distinct large orange dots
– remarkably larger than those in the non-fraudulent globules. These
are IMEIs being accessed from different SIMs – as many as orange
lines end in the dot.
Like the IMEI of a SIMbox, housing several SIMs.
www.wiseathena.com | [email protected] 21
How smart visualizations improve your business decisions
Revenue Assurance staff need access to data to make important
decisions and meet their objectives. This poses a challenge for
organizations because information is normally spread across multiple
locations and systems. Visual data exploration is the answer to this
challenge. It enables decision makers to explore data to find and
comprehend information in unified form.
Smart visualizations of data are the best tool to uncover enhanced
new insights where traditional methods fail.
Wise Athena understands the value delivered by visualizations,
and provides them to customers customized to suit their needs.
Following the defining principle of the system, visualizations adapt to
the operator data and not vice versa. That is why we call them smart.
There is no need for the customer to structure data in a certain way to
produce the graphical representation.
Wise Athena’s visualization technology stimulates data exploration
and discovery. Operators are provided with a big picture of their data
that opens doors to new insights.
www.wiseathena.com | [email protected] 22
www.wiseathena.com | [email protected] 23
www.wiseathena.com
Behavioral Segmentation
VIP Prospects Identification
Competitors Geolocalized
Perfomance
New Subscriber Profitability
Prediction
Cross-sell Opportunities
Fraud Identification
www.wiseathena.com
www.wiseathena.com
www.wiseathena.com www.wiseathena.com
www.wiseathena.comwww.wiseathena.com
www.wiseathena.com | [email protected] 24
Wise Athena offers an on-demand delivery of Telco Fraud Service Detection
on the Internet with pay-as-you-go pricing. Our system requires no setup
or maintenance. Results are produced in just four weeks compared with the
long setup periods required by traditional tools.
Business Resultsin Four Weeks
wise athena
How it works for you
With our Software as a Service (SaaS) product, all we need is your
data. Simply use an intuitive interface to upload your files.
Wise Athena automatically processes the information and performs
the analysis needed to locate and neutralize the SIMbox fraud
schemes that are taking advantage of your network.
Our team of expert Data Scientists and Business Analysts take care of
all the work required for the system to produce the relevant results
for your organization. And they also make sure that these results are
fully actionable.
Wise Athena delivers all the information the operator needs to tackle
its SIMbox fraud problems and also the smart visualizations of data
that allow for further analysis of the network and its behavior.
The use of machine learning and the fact that the tools are specifically
designed for Telco fraud mean that the data needed to accurately
train the algorithm and produce results is significantly smaller than
other methods.
You will be able to support your Revenue Assurance staff and
executive decision-makers; only paying for revenue-generating
services consumed on a variable basis.
www.wiseathena.com | [email protected] 25
It only gets better with time
Wise Athena’s fraud detection service is trained to learn from the
interactions and multiple outcomes of your network. The system finds
and evaluates relationships in large volumes of data-points. It builds a
solid data-backed foundation for crucial fraud decision-making.
Benefits of the SaaS model in the cloud reduce your overall fraud-
detection costs in multiple ways. With our service you benefit from:
• Economies of scale
• A team of leading Data Scientists and Business Analysts working
for your organization
• Efficiency improvements to continually lower prices
• A tenth of the cost and a sixth of the time required by other
providers
• No costly, time-consuming, upfront infrastructure investment
• No order, delivery, installation, and configuration of software
• No training your staff on yet more new software
As Wise Athena drives down up-front and ongoing IT labor costs, its
algorithms keep learning to provide ever-updated results.
This way you have access to a highly distributed, full-featured
platform that always yields current and relevant results – at a fraction
of the cost of traditional infrastructure and software.
With Wise Athena Cloud Computing Fraud Service you can finally
shift your valuable resources away from data center investments and
operations, into your innovative, revenue growing new projects while
you hold fraud in check.
www.wiseathena.com | [email protected] 26
Your savings with Wise Athena
SIMbox fraudsters’ business model breaks even at twenty-three
minutes of call time. Currently operators can detect the fraud at 134
minutes of use – but by this point SIMbox fraud has already been
profitable for fraudsters.
Wise Athena enables detection at seventeen minutes of call time
from the first processing iteration.
For this early detection we assume an eleven-minute delay in
accessing the data, and that the SIM credit is set to zero.
The system’s self-learning capability ensures a quick reaction to
fraudsters’ new methods of concealing their activities. It further
reduces detection time every iteration.
On these premises, the fraudsters’ business case is completely
neutralized.
And we provide these results with an accuracy of 0.0001% false
positives where operators have 1% false positives.
In other words, one ten-thousandth fewer chances of mistaking a
legitimate SIM for a fraudulent one.
Our platform is operational 24x7 and it returns results in under one
minute. You can practically see the fraud happening in real time.
www.wiseathena.com | [email protected] 27
www.wiseathena.com | [email protected] 28
In just four weeks, you will see that fraud is not an unavoidable –and
uncomfortable- part of running a business. To the contrary, fraud is
something that you can locate, isolate and get rid of once and for all.
Because if it changes form, the self-learning system will be able to find
it before it reaches profitability. And the more the data, the better the
results so your system will be more secure for your customers and more
hostile for fraudsters.
The next step for your fraud is to stop.
Just upload your network data to our cognitive analytics system and see
near real-time actionable results coming back to your decision-makers.
The next step for your fraud
wise athena
Wise Athena was assembled around one
common belief: the traditional path that
led from data to new revenues was overly
complicated and ineffective. Our pursuit of a
better way laid the foundation for this company.
Our team is made up of highly specialized
data scientists with a track record
in delivering business results and
technological excellence across a wide
range of industry sectors. We come
from Stanford, MIT, Wharton, IBM,
PriceWaterhouseCoopers, HP, Telefónica,
the Higgs Boson research team and several
start-ups.
Our combined experience comprises
the delivery of cloud services processing
more than 150M customers’ data and the
management of systems supporting up to
90M customers.
Other services offered by Wise Athena:
• CUSTOMER BEHAVIORAL ANALYTICS
• CHURN PREDICTION
• DEALER PERFORMANCE
• IDENTIFICATION OF VIP PROSPECTS
Contact us:
Wise Athena USA
Suite 400
71 Stevenson Street
San Francisco
California
Email: [email protected]
Wise Athena EUROPE Oficina 14
Av. de la Victoria 25
28023 Madrid
Email: [email protected]
About Wise Athena
The Telco industry needs to move on to a predictive model. Don’t stop at detecting existing fraud.
Be proactive: create a deterrent for the future, andneutralize the fraud business.