Nobody will disagree that in the world today the most prominent buzz word in technology is the Internet of Things (IoT) and IoT in a way will be a revolutionary factor for the transformations in almost every aspect of human lives - be it home activities, transportation, industrial operations, medical world and what not. IoT can be considered as the network of the connected physical 'things' that are accessible via internet. The term 'thing' in IoT is any device you can think of and that can be fitted with sensors, controller, compute functionality and communication capabilities. It could be your refrigerator, washing machine, car etc. As IoT enables these so called things to be accessed through Internet they are assigned unique IP addresses to transmit and receive data over the network without any manual intervention. Compute capability along with various sensors interfaces in these things allow them to interact with the external environment, collecting and transmitting data which is finally used for decision making and or analysis. Below is the pictorial representation of IoT-centric system

The IoT devices are in a way trained in such a manner that they keep sending their data consistently to the target destinations for remote data analysis. However, this sounds simple but there are some problems which are faced during the management and analysis of the data sent by IoT devices. This is because of the limited resources both for storing and computing the huge data generated by IoT devices. Some of the issues that need to be addressed while managing and analysing the IoT data are discussed below:

1. Data Quality: IoT devices keep generating massive amount of data at any given point of time. But the challenge which is faced is that sometimes the quality of the data got compromised due to numerous factors such as sensors are sending bad data, devices go offline, some errors arise in integration platform, etc. These problems may be handled by employing automated tools to a large degree. This includes identifying data quality problems, determining their seriousness, and fixing them both after data have been collected and at the IoT node.
2. Multiple Data Sources & Types: IoT systems need to employ faster curators for multiple data sources. Data curator allows systems to keep track of their data sources, their formats, and their inter relationships. The Data curators will have to make widespread usage of tools like machine learning, which are already being applied for data curation by some companies and vendors.
3. Storing IoT Data: IoT devices generate massive amount of data and in order to use and analyse it in future it is required to be stored in whatever format it comes. To perform this, a large Big Data based storage can be formed either on the local network or on cloud so that data is available when required.
4. Predictive Analytics: Mostly the descriptive analysis is applied on IoT data by generating bar charts, alerts, means and medians. But these analyses are not as fruitful as predictive analytics. For example, one would be more interested in knowing that when his machine will break down, whether the train is likely to arrive on time or not, etc.
5. Automated Decision Systems: IoT data keep flowing to its destination at very fast pace and it is not possible to assign human force to keep monitoring and analysing it. Thus, it is important to have some automated tools connected with the Big data pool that keep accessing the data from it, remove the redundancy and analyse it to generate some meaningful information.
6. Machine Learning Enabled Analytical Models: As discussed it is not possible to analyse the big data flowing in with very fast pace using traditional hypothesis. Machine learning can be a very good tool to generate and built automated analytical models to perform such tasks. Machine learning models can also be helpful in identifying unauthorized intruders into the systems, which is critical for IoT security.
7. Deep Learning Models for Image and Sound Data: Deep learning, which is based on neural network methods, is the best way to analyze large amounts of image and sound data. Deep learning models are the way to make sense of this data. They can also be used to identify patterns in cybersecurity attacks.

By - Dr. Sapna Saxena - Associate Professor, Department of Computer Science and Engineering