43
Internet of Things and Artificial Intelligence
3.1.2 The Internet of Insecure Things
IoT technology is undergoing incredibly fast development and is a heterogeneous network of small yet lightweight devices; security aspects are to be addressed for IoT deployment so that it can suitably address the issues that may come in the way while used for
both personal and commercial purposes (Eijndhoven, 2016). To that effect, the networking
capability is incorporated into the IoT devices with proper encryption; firewalls and antivirus methods are applied to counter the security concerns. This way, the confidentiality
of the data and privacy of the users are guaranteed. But, this poses some challenges also.
For example, in our houses, we usually have a physical means of access control to decide
on who can access our house and similarly we have physical means of authentication in
deciding who can enter into our house and can use the available artifacts inside the house.
Suppose we want to turn on the air conditioner in our house to make the room cool. Here,
two possibilities arise: one is to do it by physically being present in the house and following the procedures to switch on the AC. The other is, if we are away from home, to switch
on the AC before we reach home so that the room is already cool by the time we reach
home. In the second case, one is physically not present to switch on the AC. In order to
switch on the AC, even though one is physically not present in the house, the AC should
be connected as networked objects to have access control, which needs an extra layer of
security in comparison to the early nonnetworked objects concepts, for preventing any
unauthorized or malicious access to such devices.
The rest of the chapter is organized as follows. Section 3.2 discusses artificial Intelligence
(AI) basics followed by IoT basics in Section 3.3. The fusion of IoT and AI is discussed
in Section 3.4. Section 3.5 discusses the implication of deep learning in IoT scenario.
While Section 3.6 discusses the proposed methodology, Section 3.7 presents the results
and discussion. Finally, we conclude in Section 3.8.
3.2 Artificial intelligence
As we move toward the highly connected digital world, everything or anything goes smart
with the use of small lightweight sensors with distributed intelligence; a huge amount of
data is being collected from such networked devices, which poses real-time challenges in
dealing with them for better insights and making corrective action thereto. The data is so
BIG a data that even if one takes a sample of it for processing, time and accuracy becomes a
challenge. For example, in the case of wearable computing, where sensors are implanted in
the human body and are interconnected and connected to the Internet, any health-related
issues arising from the patient’s body can immediately be send to the concerned doctor for
taking the necessary action. This type of real-time processing poses challenges to IoT for
its effective and efficient implementation. Here, IoT combined with AI may be thought of
as a viable solution to address the issue at large and help to uncover the hidden information from the data, for intelligent decision-making (Aadhityan, 2015).
3.2.1 Machine Learning
Machine learning introduced in 1950 is considered to be a technique for AI, initially
aimed at robust and viable algorithms for numerous applications such as bioinformatics, intrusion detection, spam detection, forecasting, and the smart grid to name a few.
Machine learning is a powerful tool for analyzing IoT datasets (Xu, 2015).
Internet of Things and Artificial Intelligence
3.1.2 The Internet of Insecure Things
IoT technology is undergoing incredibly fast development and is a heterogeneous network of small yet lightweight devices; security aspects are to be addressed for IoT deployment so that it can suitably address the issues that may come in the way while used for
both personal and commercial purposes (Eijndhoven, 2016). To that effect, the networking
capability is incorporated into the IoT devices with proper encryption; firewalls and antivirus methods are applied to counter the security concerns. This way, the confidentiality
of the data and privacy of the users are guaranteed. But, this poses some challenges also.
For example, in our houses, we usually have a physical means of access control to decide
on who can access our house and similarly we have physical means of authentication in
deciding who can enter into our house and can use the available artifacts inside the house.
Suppose we want to turn on the air conditioner in our house to make the room cool. Here,
two possibilities arise: one is to do it by physically being present in the house and following the procedures to switch on the AC. The other is, if we are away from home, to switch
on the AC before we reach home so that the room is already cool by the time we reach
home. In the second case, one is physically not present to switch on the AC. In order to
switch on the AC, even though one is physically not present in the house, the AC should
be connected as networked objects to have access control, which needs an extra layer of
security in comparison to the early nonnetworked objects concepts, for preventing any
unauthorized or malicious access to such devices.
The rest of the chapter is organized as follows. Section 3.2 discusses artificial Intelligence
(AI) basics followed by IoT basics in Section 3.3. The fusion of IoT and AI is discussed
in Section 3.4. Section 3.5 discusses the implication of deep learning in IoT scenario.
While Section 3.6 discusses the proposed methodology, Section 3.7 presents the results
and discussion. Finally, we conclude in Section 3.8.
3.2 Artificial intelligence
As we move toward the highly connected digital world, everything or anything goes smart
with the use of small lightweight sensors with distributed intelligence; a huge amount of
data is being collected from such networked devices, which poses real-time challenges in
dealing with them for better insights and making corrective action thereto. The data is so
BIG a data that even if one takes a sample of it for processing, time and accuracy becomes a
challenge. For example, in the case of wearable computing, where sensors are implanted in
the human body and are interconnected and connected to the Internet, any health-related
issues arising from the patient’s body can immediately be send to the concerned doctor for
taking the necessary action. This type of real-time processing poses challenges to IoT for
its effective and efficient implementation. Here, IoT combined with AI may be thought of
as a viable solution to address the issue at large and help to uncover the hidden information from the data, for intelligent decision-making (Aadhityan, 2015).
3.2.1 Machine Learning
Machine learning introduced in 1950 is considered to be a technique for AI, initially
aimed at robust and viable algorithms for numerous applications such as bioinformatics, intrusion detection, spam detection, forecasting, and the smart grid to name a few.
Machine learning is a powerful tool for analyzing IoT datasets (Xu, 2015).