Machine-to-machine communications

Machine-to-Machine (M2M) communication is the next-generation telemetry which is used for automatic transmission of data gathered from remote sensors to a central unit for analysis, either by human beings or software agents. Unlike traditional Human-to-Human (H2H) communication, the human is not the typical initiator of the communication process. That is, the human is merely the recipient and possibly the respondent for the output. In contrast to conventional telemetry, M2M encompasses a broad spectrum of applications rather than just relegated to highly esoteric applications such as aerospace, water treatment and natural gas pipeline monitoring. Furthermore, M2M communications systems are composed of a myriad of machines that are connected to the Internet using public fixed and/or wireless communications infrastructure. Latest commercial forecasts are for fifty billion machines connected to the Internet worldwide by the end of the decade.

A machine-to-machine (M2M) communications eco-system is a large-scale network with diverse applications and a massive number of interconnected heterogeneous machines (e.g., sensors, vending machines and vehicles). Cellular wireless technologies will be a potential candidate for providing the last mile M2M connectivity. Thus, the Third-Generation Partnership Project (3GPP) and IEEE 802.16p, have both specified an overall cellular M2M reference architecture. The European Telecommunications Standards Institute (ETSI), in contrast, has defined a service- oriented M2M architecture. This article reviews and compares the three architectures. As a result, the 3GPP and 802.16p M2M architectures, which are functionally equivalent, complement the ETSI one. Therefore, we propose to combine the ETSI and 3GPP architectures, yielding a cellular-centric M2M service architecture. Our proposed architecture advocates the use of M2M relay nodes as a data concentrator. 

The M2M relay implements a tunnel-based aggregation scheme which coalesces data from several machines destined to the same tunnel exit-point. The aggregation scheme is also employed at the M2M gateway and the cellular base station. Numerical results show a significant reduction in protocol overheads as compared to not using aggregation at the expense of packet delay. However, the delay rapidly decreases with increasing machine density.

Let’s discuss one by one start with the underlining communications

Machine-to-machine (M2M) communication allows machines and devices to pass along small amounts of information to other machines. This includes communication to and from smoke detectors, door locks, alarms, water meters, agricultural sensors, smart buildings, smart lighting, environmental sensors, and more. Every IoT application has a different set of constraints in terms of wireless range and energy consumption it needs to achieve. Therefore, M2M network architecture is about properly utilizing radio resources. Each network listed below utilizes a different method for handling these resources. Cellular, for instance, is the only type of ubiquitous M2M network that uses its own licensed frequency space. The rest typically coexist using free, unlicensed frequencies. Due to regulatory constraints, companies are not allowed to design their networks to have an unfair advantage over other networks, so the question for these companies when creating network architecture is how to utilize the unlicensed spectrum efficiently.

Below, we’ll walk through the benefits and considerations of a few M2M network architectures currently in use. As you can see, there are many IoT networks available. Each of them is trying a unique approach to solve a standard engineering problem: how to trade off cost, performance, and complexity. Every engineer knows you can’t have the best of all of those things—but you can create a network that will cater to specific applications. We’re eager to see how these network architectures improve, evolve, and grow in the coming years. 

Cellular communication (communication based on communicating thru cellular network) has dominated the M2M space for a long time. The primary benefit of cellular is the ubiquitous coverage, but major disadvantages of cellular are short battery life, high-cost end points, and high recurring fees. Any battery-powered application will have a hard time using a cell modem. Cellular networks are constantly changing, as well. For example, when M2M started, most of the cellular world was using GSM-based technology (which is now being phased out). GSM has mostly been replaced by 3G and LTE, and there is talk that those technologies for M2M applications will eventually be phased out and replaced by LTE-M. So, companies who deployed cellular modems should be aware that their hardware may not be supported in coming years.

Great way to understand this please visit AT &T program for IoT enthusiasts with industrial application started to rollout in late sixteens https://starterkit.att.com/

 

WiFi has become a more prevalent M2M option in the last five years. This is due in part to new WiFi chip manufacturers who are now targeting the space by making lower cost, lower power chip sets with a very simple interface. With these new chips, you don’t need a computer and a WiFi driver; you can use a universal asynchronous receiver/transceiver (UART) instead.  But while cellular coverage is ubiquitous, WiFi coverage is not, which is one of WiFi’s main downfalls in the M2M market. For example, if you’re building a keycard door lock for every apartment in a New York high-rise and using WiFi, provisioning is going to be a nightmare.

Bluetooth, this option that’s become available in the last four years is Bluetooth Low Energy (BLE), which is also called Bluetooth 4.0 or Bluetooth Smart. BLE uses considerably less power than traditional Bluetooth, but like its predecessor, users are pretty limited by range and packet sizes. BLE is meant to transmit only very small bits of information online through a phone or computer. That makes BLE ideal for applications like heart rate monitors or fitness trackers, but it’s not ideal for anything that needs a stronger power draw or wider range.

ZigBee is a mesh network protocol that is trying to solve the issue of range. While it offers considerably better range than something like BLE, there are range constraints and downfalls that come with the mesh network. For example, some of the nodes in a mesh network are there just to relay information, which causes a constant (and somewhat unnecessary) power draw. This makes ZigBee a bad candidate for battery-powered devices but good for something like electric grid monitoring, which has an unlimited power source. In short, ZigBee continues to be adopted by some niche markets, but it won’t meet the needs of everyone in the M2M space.

The low power, wide-area network (LPWAN) space has recently become more saturated—and right now the leader in the group is SIGFOX. This M2M network sends small, slow bursts of data, which makes it ideal for things like alarm systems or simple meters. Due to its asymmetric link budget, the network only allows for limited r bi-directionality, so it isn’t able to send data back from the gateway to nodes at the fringes of the network. (This is a problem other LPWAN players are looking to solve.)

LoRaWAN is the M2M protocol created by the LoRa Alliance to create an ecosystem of M2M applications all using the LoRa physical layer. Like SIGFOX, LoRaWAN is an uplink-focused network and thus works well for sensor-based devices. This is partially due to regulations in Europe, which hold every device (including the gateway) to a 1% duty cycle. Because of the regulatory differences here in the U.S., a big segment of the market can be addressed by designing a protocol that allows more “command and control”-based applications. And that’s where we at Link Labs have tried to put our focus.

Symphony Link is the IoT network we at Link Labs developed in an effort to solve some of the challenges presented by other M2M architectures. For instance, a single Symphony gateway can be used to talk to 10,000 nodes, and thus cover an entire building. Symphony also targets battery life; a node on our network that sends a message every 10 minutes could feasibly last between eight to 10 years depending on the application.

this just the first part of IoT communication many to come as markets evolve. Feel free to contact me at ravindrapande@gmail.com. I would like to understand if I am missing some important angle in this technology & your view on my writing as well. 

Pattern recognition machine learning

Our researchers in artificial intelligence are harnessing the explosion of digital data and computational power with advanced algorithms to enable collaborative and natural interactions between people and machines that extend the human ability to sense, learn and understand. The research infuses computers, materials and systems with the ability to reason, communicate and perform with humanlike skill and agility.

Pattern recognition is a branch of machine learning that focuses on the recognition of patterns and regularities in data, although it is in some cases considered to be nearly synonymous with machine learning. Pattern recognition systems are in many cases trained from labeled "training" data (supervised learning), but when no labeled data are available other algorithms can be used to discover previously unknown patterns (unsupervised learning).

The terms pattern recognition, machine learning, data mining and knowledge Discovery in Databases (KDD) are hard to separate, as they largely overlap in their scope. Machine learning is the common term for supervised learning methods and originates from artificial intelligence, whereas KDD and data mining have a larger focus on unsupervised methods and stronger connection to business use.

Pattern recognition has its origins in engineering, whereas machine learning grew out of computer science. However, these activities can be viewed as two facets of the same field, and together they have undergone substantial development over the past ten years. In particular, Bayesian methods have grown from a specialist niche to become mainstream, while graphical models have emerged as a general framework for describing and applying probabilistic models. Also, the practical applicability of Bayesian methods has been greatly enhanced through the development of a range of approximate inference algorithms such as variational Bayes and expectation propagation. Similarly, new models based on kernels have had significant impact on both algorithms and applications.

The problem of searching for patterns in data is a fundamental one and has a long and successful history. For instance, the extensive astronomical observations of Tycho Brahe in the 16th century allowed Johannes Kepler to discover the empirical laws of planetary motion, which in turn provided a springboard for the development of classical mechanics. Similarly, the discovery of regularities in atomic spectra played a key role in the development and verification of quantum physics in the early twentieth

century. The field of pattern recognition is concerned with the automatic discovery of regularities in data through the use of computer algorithms and with the use of these regularities to take actions such as classifying the data into different categories. Consider the simple example of recognizing handwritten digits, each digit corresponds to a 28×28 pixel image and so can be represented by a vector

x comprising 784 real numbers. The goal is to build a machine that will take such a vector x as input and that will produce the identity of the digit 0, . . . , 9 as the output. This is a nontrivial problem due to the wide variability of handwriting. It could be tackled using handcrafted rules or heuristics for distinguishing the digits based on the shapes of the strokes, but in practice such an approach leads to a proliferation of rules and of exceptions to the rules and so on, and invariably gives poor results.

Far better results can be obtained by adopting a machine learning approach in which a large set of N digits {x1, . . . , xN} called a training set is used to tune the parameters of an adaptive model. The categories of the digits in the training set are known in advance, typically by inspecting them individually and hand-labelling them. We can express the category of a digit using target vector t, which represents the identity of the corresponding digit. Suitable techniques for representing categories in terms of vectors will be discussed later. Note that there is one such target vector t for each digit image x.

-The result of running the machine learning algorithm can be expressed as a function y(x) which takes a new digit image x as input and that generates an output vector y, encoded in the same way as the target vectors. The precise form of the function y(x) is determined during the training phase, also known as the learning phase, on the basis of the training data. Once the model is trained it can then determine the identity of new digit images, which are said to comprise a test set. The ability to categorize correctly new examples that differ from those used for training is known as generalization.

In practical applications, the variability of the input vectors will be such that the training data can comprise only a tiny fraction of all

possible input vectors, and so generalization is a central goal in pattern recognition.

For most practical applications, the original input variables are typically preprocessed to transform them into some new space of variables where, it is hoped, the pattern recognition problem will be easier to solve. For instance, in the digit recognition

problem, the images of the digits are typically translated and scaled so that each digit is contained within a box of a fixed size. This greatly reduces the variability within each digit class, because the location and scale of all the digits are now the same, which makes it much easier for a subsequent pattern recognition algorithm to distinguish between the different classes. This pre-processing stage is sometimes also called feature extraction. Note that new test data must be pre-processed using the same steps as the training data.

Pre-processing might also be performed in order to speed up computation. For example, if the goal is real-time face detection in a high-resolution video stream, the computer must handle huge numbers of pixels per second, and presenting these directly to a complex pattern recognition algorithm may be computationally infeasible.

Instead, the aim is to find useful features that are fast to compute, and yet that  also preserve useful discriminatory information enabling faces to be distinguished from non-faces. These features are then used as the inputs to the pattern recognition algorithm. For instance, the average value of the image intensity over a rectangular sub-region can be evaluated extremely efficiently, and a set of such features can prove very effective in fast face detection. Because the number of such features is smaller than the number of pixels, this kind of pre-processing represents a form of dimensionality reduction. Care must be taken during pre-processing because often information is discarded, and if this information is important to the solution of the problem then the overall accuracy of the system can suffer.

Applications in which the training data comprises examples of the input vectors along with their corresponding target vectors are known as supervised learning problems.

Cases such as the digit recognition example, in which the aim is to assign each input vector to one of a finite number of discrete categories, are called classification problems. If the desired output consists of one or more continuous variables, then the task is called regression. An example of a regression problem would be the prediction of the yield in a chemical manufacturing process in which the inputs consist of the concentrations of reactants, the temperature, and the pressure.

In other pattern recognition problems, the training data consists of a set of input vectors x without any corresponding target values. The goal in such unsupervised learning problems may be to discover groups of similar examples within the data, where it is called clustering, or to determine the distribution of data within the input space, known as density estimation, or to project the data from a high-dimensional space down to two or three dimensions for the purpose of visualization.

Finally, the technique of reinforcement learning is concerned with the problem of finding suitable actions to take in a given situation in

order to maximize a reward. Here the learning algorithm is not given examples of optimal outputs, in contrast to supervised learning, but must instead discover them by a process of trial and error. Typically, there is a sequence of states and actions in which the learning algorithm is interacting with its environment. In many cases, the current action not only affects the immediate reward but also has an impact on the reward at all subsequent time steps.

Let’s discuss a live example, ANPR (Automatic number plate recognition) or LPR (License plate recognition)  It is the technical method of artificial vision (OCR) that allows the recognition of number plates in images of vehicles. Historically, it has been applied on security systems to control accesses of vehicles and car parks.

Nowadays, the ANPR technology has improved its reliability, some systems are able to offer recognition rates between 95 and 98%. Also, some ANPR equipment are able to recognize the number plate of vehicles that drive up tol 200km/h.

Generally, the ANPR technology can be bought in two modalities, The ANPR engine The ANPR equipment (Hardware + recognition engine)

The ANPR engine can recognize the number plate directly from the images stored in a hard disk. This type of software allows to take good use of images that have been obtained from other systems like CCTV or cameras.

The ANPR equipment incorporates all the hardware necessary to capture the images of the vehicles and to recognize the number plate. Moreover, it incorporates the ANPR engine. The ANPR equipment are designed to offer the maximum reliability.

The ANPR process is divided into three steps. The detection of the vehicle, the capture of the images and the process of recognition. Next, we will detail step by step how it works and depending on each case what the advantages and disadvantages are

Once the vehicle is detected, the following step is the capture of the vehicle. In order to take a right image, the following points will have to be considered. Each ANPR manufacturer has developed its own recognition algorithms, although, these are the main ones and the common ones

    To locate and to isolate the number plate in the image

    To correct the brightness and the contrast of the number plate

    To separate each character of the number plate

    To recognize each character of the number plate

The Capture Unit that takes the image of the vehicle, and the Process Unit that receive the image from the Capture Unit and makes the recognition of the number plate. The Process Units can control one or more Capture Units simultaneously.

The quality of the solution depends on each manufacturer, although generally the "CU+PU" architecture has disadvantages respect the ANPR All-in-One. Now let’s apply the modus we have discussed above & try to get a solution. Then you could search the solution on web & check the feasibility of your thoughts.

If you need any help feel free to get in touch with me at ravindrapande@gmail.com & I will be happy to help.

IoT Risks Fin Institutes

We've all have discussed on the Internet of Things by now, billions of devices connected to the internet, gathering all kinds of information on us and our daily lives. And while many of the attention-grabbing headlines will highlight the consumer-facing Internet of Things, such as in cars, domestic appliances and healthcare, the industrial sector is also already embracing connected devices.

In their Asia/Pacific Internet of Things Market Forecast, predicts that by 2020 there will be 8.6 billion connected devices in APeJ. Smart Grids will be the leading use case, followed by Manufacturing Operations, Asset/Fleet Management and Smart Buildings. According to the report, by 2020 the total market opportunity in the IoT ecosystem will be in excess of $500 billion, of which the anticipated spend on security will be $8 billion plus.

Utilities, energy providers and manufacturers are increasingly looking to connected devices to help their business streamline industrial control systems (ICS). But just as there are worries over IoT security, ICS are also facing increasing security threats, and connected devices further highlights the need for proper security measures. According to a report, Asia-Pacific Industrial Control Systems Security Market, the APAC market for ICS security is set to top $1 billion in just four years as industry players begin to understand the growing cyber threat to operational technology.

What we're addressing here is the Industrial Internet of Things, and attacks on it are already fairly common. This is a particular worry because the very foundation of IoT, and indeed IIoT - what makes it such a game changer - is also its security weak spot. By this we mean the fact that all these different components - typically manufactured by different vendors - talk to each other. And these vendors can and do require remote access to systems for a variety of reasons, such as pushing out updates or collecting data. And because many of these vendors originally come from the consumer sector, security perhaps isn't built into their devices as much as it should be. PCI have mentioned specific concerns over the IOT platform add hoc extensions. The security risks are quite major concern in Financial institutions as well.

So IIoT vendors can be targeted by cyber criminals as a way to gain access to a specific organization. It's one more route to bypass a company's defenses attacking a third party that interacts with or maintains part of the connected infrastructure of another business.
We have seen attacks like this already. This attack, for example, targeted three companies that make software for the industrial sector. Malicious code was implanted into their software update processes, which when their customers updated was transferred to their systems, giving the attackers access to vital data, systems and services.

The fact that the attackers were able to introduce malicious updates to the victims' servers strongly suggests that they had some sort of internal access to the network. It is also likely that they would have had sufficient permissions to upload the infected updates. These privileges or permissions are associated with human accounts or automated systems and if these are not properly managed, if a company loses control of critical administrative login data, for example, they can be hijacked.

Now, it is of course very difficult for a customer to have any influence over the security a vendor has put in place. But there are some things that can be done. It is vital, for example, that customers understand the dependencies within the supply chain, and where any weaknesses lie. Any links within that chain should have the same level of control that exist internally.  Also, it's possible that when working with vendors, customers can negotiate contracts or SLAs that guarantee sufficient security controls. This can be specific to interaction between the vendor and customer, such as ensuring the integrity of updates before the customer downloads them.

It's also worth considering whether there is a human element involved, and what controls are in place to ensure credentials are secure. The same process can be applied when it comes to who at the vendor has the right to access the customer environment. Credential management like this can control who has the privilege or permission to gain remote access into your infrastructure.  Ultimately, the openness of the IoT and communication among its different elements can and should be extended to include vendors and customers. That's key to ensuring your business remains secure.

Financial technology Risks has reached a tipping point, today more and more financial institutions are noticing the benefits that technology offers users, from convenient services to real-time access amid the rapid proliferation of mobile devices and cloud computing of recent years.

According to Accenture's analysis of CB Insights data, investments in APAC various ventures, primarily in China, reached almost US$10 billion as of current year July end,- more than twice the US$4 billion invested in the region in all of 2015. The top 10 investments in APAC ventures occurred in China and Hong Kong were accounting for 90% of all investments in the region.  Evidence of this growth is all around us today. For example, according to PwC's Global Economic Crime Survey 2016, the number of consumers using digital banking in Asia Pacific reached 670 million in 2014, and is expected to increase to 1.7 billion by 2020. The service has revolutionized the banking industry, leading to a growth in online and mobile banking of 35% on average annually, while the use of traditional banking decreased by more than a quarter.

Banking industry is being put at risk. Not limited to the launch of Apple Pay and the first batch of stored value facility (SVF) licenses to provide e-wallet services by the Hong Kong Monetary Authority (HKMA), they bring huge convenience to daily life, while at the same time, though the growth presents significant benefits to the industry, it also brings about significant risks. Recent incidents across Asia, and in particular in Hong Kong, have drawn attention to the security risks associated with digital banking.

HKMA recently revealed that there are at least 22 online bank accounts in at least four banks that have reported unauthorized stock trading activities, totaling a sum of HK$45.97 million. Although HKMA said that none of the cases reported resulted in any fund transfers to unregistered third parties (thanks to a double authentication process), there were nine cases that resulted in financial losses of HK$1.56 million. For the banks, the fallout extends beyond just financial liability, and could have lasting impacts on everything from consumer trust to organizational reputation.

Bring security awareness on a healthy level: According to a recent study by F5 and The Asian Banker, the majority (84%) of financial firms now rank cyber threats as one of their top business risks. CEOs are increasingly concerned about the impact of these threats on their business, but less than half (37%) of organizations actually have a cyber incident response plan or policy in place.

Threats are becoming increasingly sophisticated and creative. The five most common threats organizations face are malware, web application attacks, point of sale attacks, insider compromise and DDoS attacks. Despite this, end users are increasingly used as an alternative channel of launching attack due to the sheer number of devices, many of which are unknown - and unsecured. Awareness is growing about this and other threats, but it is a cat-and-mouse game, with criminals switching tactics and inventing new methods of attacks regularly.

Prevention is better than mitigation : Regulators are aware of this threat, and increasingly they are taking steps to mitigate the risks. The HKMA has announced the launch of a Cyber security Fortification Initiative (CFI) at the Cyber Security Summit 2016, and issued a formal circular to all banks setting out that it is a supervisory requirement for them to implement the CFI. This initiative will enhance the protection of multiple banking channels.

For banks and financial institutions strategies are needed that offer real-time threat identification, deep analysis and comprehensive protection due to the dynamic nature of their operations. They should stay vigilant and focus their effort on three items.

First, they need to prioritize real time monitoring and prevention, to guard against malware and phishing attacks which are designed to steal identity, data and money at any time. Second, they need to make sure that no endpoint software or user involvement will be required and have full transparency on the security control. Third, they also need a multi-device support, to protect transactions made on any devices or channels as every transaction can be at risk.

Cyber crime is the greatest threat that banks and financial institutions face today. Careful planning and prompt action for when, not if, organizations are threatened could mean the difference between competitive success, or financial failure.