In this study, an IoT framework with a four-layer and blockchain storage system has been developed for the secure and high-accuracy operation of IoT assets with distributed-based learning systems. In the second part of the study, firstly, an IoT device was developed for processing nodes that collect data at endpoints. This IoT end device is responsible for transmitting temperature and humidity data to the upper layer. In the second step, edge devices were developed to transfer the collected node data to the server. These edge devices are responsible for both transferring the data to the server and transmitting the learning results to the nodes. In the last step, the data collected in the center is learned with the LSTM learning algorithm and six different order classes are decided. In addition, it is explained how the processing data is stored in the blockchain. As a result, it is predicted that the proposed secure and blockchain-based IoT framework will contribute to the literature with its contribution as follows.
• It has been shown that distributed learning-based systems will provide higher learning performance by combining datasets on the central server and updating their weights.
• It has been shown that preferring edge devices in distributed IoT-based systems will use network traffic more efficiently.
• It has been observed that the transmission of data transfer with smart contract and storage in blockchain-based systems ensure data integrity and security.
• In IoT systems with high data security, solution-specific development of edge and node devices has enabled the use of faster and more secure cryptographic methods.
In the second part of the study, the technical details of the artificial intelligence-based and secure IoT infrastructure are explained. In this section, structures such as the IoT device developed, the general working architecture of the study, the prepared blockchain and smart contract structure, and the developed machine learning model are explained. In the third part, there are the findings obtained from the study. In this section, the network traffic measurements of the developed infrastructure, the speed and resource usage measurements of the infrastructure are clearly shown. In the fourth chapter, the results obtained from the study are explained.
ARTIFICIAL INTELLIGENCE BASED AND SECURED IOT FRAMEWORK
In this study, it has been tried to develop a blockchain-based distributed edge computing backbone. The study will be examined under three main headings in itself. The first part of the developed backbone consists of the technical features and visuals of the IoT card, which provides data from the environment. In the second part, the details of the designed edge computing structure and artificial intelligence processes will be discussed. In the last part, how the blockchain technology is integrated into the first and second parts will be explained.
Designed IoT Device: IoT_TH
We chose to develop the IoT card ourselves, which we will use within the layered architecture of the infrastructure we will develop. We placed sensors on the IoT card we developed to collect the temperature and humidity data of the environment. We named the card we developed using the initials of this environment data and the abbreviation of the concept of internet of things, IoT_TH. The purpose of designing this card is to measure the temperature and humidity values ​​of the environment where the card is located and to send the values ​​obtained to the remote server via the Wi-Fi module. ESP8266 is used as hardware, MCU and Wi-Fi module. DHT11 temperature and humidity sensor is used as the sensor. There is a Relay and a Buzzer on the board to activate when necessary. The reason for placing a relay on the board is to give this board the ability to control motors and similar equipment. The developed board is fed with 5V power. The relay and buzzer on the board work with 5V. The DHT11 sensor works with 3.3V, which is the same value as the ESP8266. There is 1 test pin on the PCB board to measure the total consumed current and voltage. With this pin, it is possible to measure how much power the card consumes in which situations. With the SS34 protection diode, when reverse polarity is applied to the board, damage to the circuit elements is prevented.
Under normal circumstances, when the location of the card changes or the SSID and password information of the Wi-Fi device to which it will be connected changes, the card must be reprogrammed and the SSID and password information must be reconfigured as hardware code. However, we can easily do this by dynamically adjusting the Wi-Fi configurations of the card through the switch we placed on the card. Thus, when the location of the card is changed, the information of the new Wi-Fi device to be connected is entered with the Wi-Fi configuration button and it can be easily adapted to the new location. Thanks to these features, the card has a more effective quality. The developed card is an ergonomic card with dimensions of 70mm - 30mm - 18mm. The internal structure, connection diagram and design of the developed card are shown in Figure 1.