Posts

The Prophet Jeremiah

Identifying Toll Fraud is Harder Than finding a Needle in a Haystack

What Does That Have To Do With Big Data?

[Post is better viewed on the blog Website]

The Prophet JeremiahAt the time of creation, God spoke with man directly, without any proxy. God spoke to Adam, Eve, the snake and even handled the first murder interrogation by himself when asking Cain “Where is your brother Abel?” After this, when there were too many people, God abandoned the one-on-one approach and started sending his messages and commands through prophets. Then came the kings who listened to the oracles and ignored the prophets.

And then came the scientific revolutionaries, visionaries, dreamers and most recently, the predictors. Unlike prophets, scientific revolutionaries, visionaries and the dreamers, the predictors look to the past to predict the future. And the deeper the predictor studies the past, the clearer he can envision the future.

The predictor is a by-product of an emerging technology – big data. I am sure that big data was invented by a male, since it’s totally built on a male character trait –don’t throw anything away that you may one day need.  This is probably the reason why another male invented large garages. Unlike the traditional rational data bases, big data deals with voluminous amounts of unstructured data (not organized by any method), which is gathered from many sources in large quantities, various formats and varying qualities. There are four main characteristics related to big data (aka the four Vs); Volume, Velocity, Variety, Volatility. Allow me to add a simple analogy from my life to describe the difference between rational and irrational data bases. When I return from the grocery I pile ALL the vegetables and fruit in the refrigerator inside their plastic bags. My lovely and more rational wife washes them all, skins the melons and watermelon and cuts them into pieces, peels the vegetables and sometimes cuts them as well, to be ready for making salad or cooking.

Mr. Gurdeep Singh Pall is the Corporate Vice President for Skype and Lync at Microsoft Corp. Mr. Pall just returned to the Lync unit, after spending the past two years working on Artificial Intelligence projects within Microsoft.  Pall used his opening keynote at this year’s Lync Conference to describe how the work of analytics and Bayesian predictions, will eventually make its way into communications systems. In practice, Singh Pall said, “We can actually predict who you will be calling in the next five minutes.”

Big data and Unified Communications & VoIP services

Unified Communications and VoIP services collect a lot of raw data; this data is worthy of analysis due to its wealth of intelligence. Many companies are increasingly aware that there is information that can be collected and refined to an essence which can be used for performance optimization and network design improvements. UC and VoIP big data analytics will be the key element in converting the big data to a tool which ensures cloud-based VoIP service, including privacy, security, toll fraud, performance, cost and more.

So what can you do with voice analytics?

  • VoIP analytics will build its own multi-layered picture of the network’s topology derived from the big data over time.
  • VoIP analytics will provide Network & Users Profiling
  • VoIP analytics will provide Advanced Call fraud detection and attack prediction
  • VoIP analytics will provide Advanced Multi-Dimensional Cost Analysis

Toll Fraud detection and prevention using big data analytics

Call fraud is associated with significant revenue loss and is hard to discover. I read that discovering call fraud in the masses of call records is more difficult than finding a needle in a haystack. Actually, that’s an easy problem to solve; you know how a needle looks like and by adding enough manpower to do the looking you can eventually find it.  But fraud calls are similar to legitimate calls, so if you can’t identify a fraud call, no matter how much manpower (or CPU power in this case) you put on the job it will be impossible to detect. It’s more like trying to find a specific strand of hay in a haystack.

A common approach to detect call fraud is based on examining accounts made up of several statistics that are computed over a specific period. For example, average call duration, longest call duration, and numbers of calls to particular countries might be computed over the past hour, several hours, day or several days. Account summaries can be compared to thresholds for each period, and an account whose summary exceeds a threshold can be queued and analyzed for fraud.

VoIP analytic fraud detection is designed on a statistical principle of dynamic VoIP fraud detection. The algorithm is based on Tracking Account Behavior which is able to alert or terminate the fraudulent call as it occurs. The algorithm will relay runtime & historical attributes gathered per user, group of users, sites, SIP interface and etc. The VoIP analytics create a signature of predicted usage behavior for each user/group/interface, update the statistical model with each call and score calls for fraud using predicted behavior as the baseline. When a call exceeds a predictive user signature boundary, the VoIP analytic may take actions as per the configuration.

The VoIP fraud detection analytic is built on three stages:

  1. Training – The analysis of large numbers of enterprises of various types such as: Unified Communications, Contact Center, etc. Based on this information, the VoIP analytic Fraud Detection System creates preliminary statistical information which is later segmented per the organization’s characteristics.
  2. Adaptation – Adjustment of the statistics collected in the previous stage to the specific organization. This is done by comparing in real-time the statistics to actual call activity of the organization.
  3. Test – Each call is compared against the statistical call pattern in real-time. Calls that don’t match the pattern will result in fraud alarms with the probability (confidence) grade.

 Conclusion

We are often tempted to impose the way we see things through the prism of our own life experiences on our friends and family while in actuality, what we are really doing is judging them for the way they see things.  A friend once told me that life experience is like a flashlight hanging on your back when you are going forward, in other words, useless.  That may be true. But in the case of Big Data analytics, the system’s life experience is the basis for predicting a better future.

Call Toll Fraud

46.3 Billion Reasons to Invest in Call Toll Fraud Prevention

[Post is better viewed on the blog Website]

I remember the summer of 1972 as one hell of a hot summer. We were a bunch of kids on summer vacation without much to do. Our daily routine was mainly playing football in the deserted schoolyard and searching for other “exciting” activities the rest of the day.  One day we decided that it would be much “cooler” to stay indoors in the classrooms. So we broke the lock of one of the classroom and entered. I don’t remember exactly what we did in the classroom but what I do remember is that for some reason we decided to remove the blackboard (in those days it was really black and heavy) and take it with us through the window. But we had no clue of what to do with the loot! That took place more than 40 years ago and I still can feel the bitter taste of helplessness kids standing on the schoolyard with stolen blackboard

Call Toll FraudI have cool job in AudioCodes. Together with my colleagues, we are working on next generation products. A few months ago, I came up with a “revolutionary” idea as to how to prevent toll fraud using big data technologies.  I met with our security expert and presented my idea. He gave me an “offering my condolences” look saying – who cares about toll fraud? Who needs to fraud when calls are so cheap? And I felt again like I did 40 years ago, defeated, sweating, carrying a huge blackboard, and this time alone. But I didn’t give up, I decided to check the numbers.

The CFCA 2013 Global Fraud Loss Survey

The Communications Fraud Control Association (CFCA) published a fascinating report, “the 2013 Global Fraud Loss Survey”,  taken from fraud and security experts working within the industry who are directly involved in identifying and stopping communications fraud.  Responses were received from 93 Communications Service Providers (CSPs) located throughout the industry and around the globe. The CSPs included companies both small (<1K employees) and large (100K+), and covered wireless, wireline, broadband, and narrowband service providers. The CSPs reported providing service in multiple areas including: voice, data, financial services, and content distribution.

According to the survey, the estimated 2013 Global Fraud Loss was $46.3 Billion (USD) annually which is approximately 2.09% of total telecom revenues. Here are some interesting findings from the survey.

Top 5 Fraud Methods Reported

  • Subscription Fraud
  • PBX Hacking
  • Account Take Over / Identity Theft
  • VoIP Hacking
  • Dealer Fraud

 

Top 5 Fraud Types Reported

  • Roaming Fraud
  • Wholesale Fraud
  • Premium Rate Service
  • Cable or Satellite Signal Theft
  • Hardware Reselling

Top 10 Countries from which Fraudulent Calls Originate:

Ten countries account for 35% of the originating global fraudulent calls

Top 10 Countries from which Fraudulent Calls Originate

Top 10 Countries from which Fraudulent Calls Originate (CFCA)

 Top 10 Countries where fraud terminates:

I find this  graph to be amazing as many of these top ten are relatively esoteric countries (sorry for the non-politically correct language) and are the destination of more than 40% of the fraudulent calls!

Top 10 Countries where fraud terminates

Top 10 Countries where fraud terminates (CFCA)

 Estimated Fraud Losses by Service Type (in $USD Billions)

 

This pie chart shows that calls (PSTN & VoIP) comprise of more than 50% of the fraud service types

Estimated Fraud Losses by Service Type

Estimated Fraud Losses by Service Type (in $USD Billions) (CFCA)

 

According the survey, the companies that are subject to fraud don’t report this to law enforcement for the following reasons:

  • Debt recovery pursued through civil means
  • No faith in the judicial system to administer the right punishment to deter others
  • No perceived value to the business
  • Not referred due to lack of evidence
  • Perceived lack of interest by law enforcement to take the case
  • Perceived lack of understanding by law enforcement to pursue the case
  • Lack of resources

 

Conclusion

According to this survey, losses are huge and the fraud trend is definitely on the rise. From estimated total global revenues of $2.214 trillion (USD) in 2013, the estimated loss due to fraud is $46.3 Billion (USD), or 2.09%. The estimated total global revenue has been growing by 3.7% since 2011 where the Estimated Global Fraud Loss is growing by 15.4%!

So what do I hope are the main takeaways from this blog post?

Firstly, don’t write posts which can incriminate you or that you don’t want your kids to read. And, don’t break into school classrooms! And now seriously… Don’t give up on your ideas, even if the “experts” say they aren’t worth anything. Fraud detection applications are absolutely viable, According to the survey, toll fraud doesn’t occur often, but when it does hit, it can be financially painful.  And what’s amazing is that CSPs and other organizations prefer not to report to law enforcement since they don’t want to be tagged as low security firms. Additionally, they also lack the confidence that there is a real chance to expose the “faceless” net criminals.

Stay tuned for my next post on how to prevent toll fraud using big data tools.

Big Data in the Service of Brain Manipulation

Big Data in the Service of Brain Manipulation

[Post is better viewed on the blog Website]

Editor’s notes: As the editor of AudioCodes Voice Blog I’m always looking for interesting, off-topic, technology posts. This post by Yossi Zadah is a good example. The post is not about VoIP, yet it is about voice and video and how our brain unconsciously makes decisions based on these elements. If you have a topic you think would be of interest to our blog readers, please contact me. We are always happy to accept guest posts. Amir Zmora.

Mental firewalls, mind control & big dataBig Data in the Service of Brain Manipulation

One of the more recent and very popular presentations on TED, was actually a playlist of 11 presentations about data privacy, entitled “The Dark Side of Data”.

In a fascinating talk by Alessandro Acquisti, named “Why Privacy Matters”, Mr. Acquisti explored the behavioral economics of privacy (and information security) in social networks. What motivates you to share your personal information online, he asked.

His team’s surprising study on facial recognition software showed that it can connect an anonymous human face to an online name – and then to a Facebook account — in about 3 seconds.

In this talk, Alessandro illustrates that any personal information can be sensitive information. In one of many of the presentation highlights, Alessandro draws a scenario of using public social media data for personalized advertisements.  He describes a futuristic method of personalized advertisements where the sophisticated marketer uses public social media data (i.e. Facebook) and by using a relativity simple algorithm, chooses two pictures of your best friends. By using a facial composite tool, the marketer creates a new picture which is in essence a combination of the two pictures of your best friends. In the next step the marketer creates a customized ad using the composite picture.  Studies show that people can’t even recognize themselves in facial composites, but they do react to them in a positive manner.

Facial Composite for Brain Manipulation

Voice

Northeastern University computer science professor Rupal Patel looks for ways to give voice to the voiceless. As founder and director of the Communication Analysis and Design Laboratory (CadLab), she developed a technology that combines real human voices with the characteristics of individual speech patterns. The result is VocaliD, an innovation that gives people who can’t speak the ability to communicate in a voice all their own.

To build custom crafted voices, Professor Patel extracts properties from a target speaker’s disordered speech (whatever sounds the target speaker can produce) and applies these features to a synthetic voice that was created from a surrogate voice donor that resembles the target speaker in age, size, sex, etc. The result is a synthetic voice that contains as much of the vocal identity of the target speaker as possible, and the speech clarity of the surrogate voice donor.

VocaliD aim is creating a worldwide surrogate voice donor database to be able to synthesize target voices as close as possible to the target voices.

Let’s try to bring together Professor Patel’s fascinating and novel way to give voice to the voiceless with Professor  Acquisti’s sophisticated futuristic advertisement method and try to extend it to other fields in our everyday lives. In this extended method, the algorithm will use public social media sources to retrieve the voices of two of your best friends.. It can be done by using video clips, various voice recordings and the like.

In the next step, the sophisticated marketer will add to the composite picture a synthesized voice which resembles that of your best friend or mother, father, teacher or any other authoritative figure in your life. Now we have a picture and voice which can be used for various purposes, not only for advertisements.

A determent agent will conduct deep data mining through public (and sometimes non-public) data in order to identify pleasant memories, sites, events, locations.  Using Computer-Generated Imagery (CGI), we can combine the facial composites, synthesized voices and fictional setting which will unconsciously direct you to what to buy, who to elect and how to react to various stimuli.

We began this journey with a facial recognition software which connects an anonymous human face to an online name and we ended-up with a mind control tool which breaches your mental firewall. The impact of this exercise can be significant as it may have dramatic effects on the advertisement industry all the way to homeland security issues.