One more business-oriented view on algorithms

In this article we would like to discuss some of the business significance of artificial intelligence or how complex analytics tools help data scientists make sense of large amounts of information whether it is historical, transactional or machine-generated.

When used properly, these algorithms help detect previously undetectable patterns for example customer buying behavior or similarities between allegedly different cyber attacks.

One big problem related to algorithms is their short life span. Besides great math, one also needs to be constantly aware which algorithms are becoming obsolete and replace them with new better ones. This needs to be done continuously and quickly as in today’s complex business environment only the most high-yielding algorithms will survive.

In a hacker attack the defense system is updated to neutralize the threat. But hackers always come back with different approach. In order to stay secure, companies have to develop algorithms at least as quickly as the cybercriminals. On Wall Street even the best algorithms are profitable for 6 weeks at the most. In this period competitors are able to reverse-engineer and exploit it. Algorithmic efficiency is what’s most important in cybersecurity. One well-known case is with a company named Target – it’s systems were able to detect the hack but they had no algorithm for separating the real hacking from unimportant errors as the spewed information was simply too voluminous.

In the financial services stakes are extremely high so algorithmic fraud detection is also highly developed. Knight Capital, an American global financial services company, lost approximately $440 in 45 minutes back in 2012. Needless to say, this company is non-existent, whereas algorithm innovation only becomes larger.

It’s essential to come up with a way of easily compiling data with a sieve for enormous amounts of irrelevant information. As systems present new patterns of failure, one can improve on what was already done with an innovative approach.

Another field that shows rapid algorithmic development is predictive maintenance. Internet of Things is expected to be part of our household items by 2020, but is already seeing industrial use. Algorithms are developed to identify initial signs of system failure and alert the command center.

There are a number of tools that help in the development of algorithms:

  • visual analytics – relates to pattern recognition being used in real time to explore enormous amounts of historical data and connect usable models
  • streaming analytics – inserting algorithms directly into streaming data in order to monitor it live and isolate patterns or detect potential threats
  • predictive analytics networks – a place where data scientists help each other and polish their algorithms with a degree of reciprocity
  • continuous streaming data marts – used to monitor an algorithm in real time with the possibility of calibrating it

Top companies are continuously improving on their algorithms. In an environment they created, only the fittest algorithms survive and they will drive smarter and sounder business choices.


Image source

Some major challenges the IoT will bring

In our previous posts we mentioned predictions for around 20 billion IoT devices connected by 2020. Today this forecast may even be pessimistic, since the Business Insider quotes numbers above 30 billion. There will be an enormous opportunity for better energy efficiency and data security. New challenges regarding the rapid growth of IoT have to be quickly understood as they might drastically slow the implementation process.

In this article we will examine some of these challenges.

Device authentication

One very important feature of IoT ecosystem security is identifying devices and thus preventing “outsiders” from entering it. Today authentication is done on cloud-based servers, proving a reliable choice when there are tens of devices connected. However, if you pile up thousands or millions of IoT devices, authentication can become a liability. In terms of security we are not advanced enough as current practices include Internet connection which drains batteries and is totally useless when the connection drops. On top of that, people in general do not understand the issue with scaling, according to Ken Tola, the CEO of the IoT security startup Phantom. He says that working on a peer-base could easily handle big scalability. This means moving functionality between IoT devices – whenever authentication is necessary, it can happen at the same time between millions of devices without requiring Internet connection.

The same startup is working on M2M (machine-to-machine) connection which is a security layer able to identify two devices in peer-based manner, identifying levels and types of communication between them.

Wireless communications

IoT is the logical expansion of the Internet from our computers to our appliances. It will digitize some of our everyday activities via wireless connectivity. Majority of IoT devices depend on radio frequencies such as Bluetooth and Wi-Fi. However, RF-based devices are shutting each other down due to interference and this problem will only grow with the addition of other IoT appliances. One current solution can consist of an additional bandwidth of 5GHz for Wi-Fi, but as projected number of devices grow more and more, interference will persist. Another issue relates to power consumption, as IoT items use batteries.

An alternative is instead of using Near Field Magnetic Induction (NFMI) for data transfer to be substituted with RF.NMFI, whose signal decays much faster and thus much of the interference is gone. NFMI creates a wireless “bubble” in which IoT devices connect and outside signals are ignored. Also, security protocols are active within the bubble, drastically reducing threats, while fast signal decay allows the same frequency to be used for a different device. On a side note, NFMI has been used in hearing aids and pacemakers for more than a decade, but it might be the key to revolutionizing IoT.

Traffic administration

Managing IoT devices can quickly become impossible if it is not taken into account at the earliest stages of implementation. Smart homes are one thing, but we will live in smart cities where parking automates or traffic sensors will also transmit data. Administration and integration should be as simple as possible. The biggest potential issue is many devices transmitting data at the same time, but as we mentioned, RF.NMFI might be the key.

Startups use machine learning (Artificial Intelligence) for managing complex automated networks that will consist of thousands IoT devices. The algorithm will provide real-time distribution system control and self-managing means for big networks,geographically spread over vast distances.

Most of the technologies we use will further develop to a level supporting the needs of an interconnected world. Those that are left behind will be replaced by ones ready to take on the challenges and opportunities the IoT will bring.


Image source

We Need To Get The Internet Of Things Right

Here is a good article from Dave Evans giving us some food for thought of how important it is to get things right with the Internet of Things (IoT). Now.

“Having grandiose visions for IoT is one thing, but IoT really does have the power to transform even mundane things into something remarkable … We must get to the point where technology works for people, rather than people working for technology. IoT is here, but our work is just beginning.”