Hardware Limitations of Lightweight Cryptographic Designs for IoT in Healthcare

: Security is an important aspect of healthcare applications that employ Internet of Things (IoT) technology. More speciﬁcally, providing privacy and ensuring the conﬁdentiality, integrity and authenticity of IoT-based designs are crucial in the health domain because the collected data are sensitive, and the continuous availability of the system is critical for the user’s wellbeing. However, the IoT consists of resource-constrained devices that increase the difﬁculty of implementing high-level-security schemes. Therefore, in the current paper, renowned lightweight cryptographic primitives and their most recent architecture, to the best of the authors’ knowledge, are investigated. Their security, architecture characteristics and overall hardware limitations are analyzed and collected in tables. Finally, all the algorithms are compared based on their effectiveness in securing healthcare applications, the utilized device and the overall implementation efﬁciency.


Introduction
The Internet of Things (IoT) is a network of heterogenous devices that are interconnected with each other and can transmit data to the Internet [1]. The utilized devices are resource constrained with computational and energy limits. In recent years, IoT has been adopted by the healthcare domain because of the new services and extensive capabilities it provides, while being low cost and easily accessible. Nonetheless, security requirements for applications in IoT-based healthcare are also expanding [2]. The design and implementation of security mechanisms are essential in smart health because a malicious attack or a device malfunction can negatively affect the users/patients, even endangering their lives. Common cryptographic algorithms that can potentially secure such systems cannot be employed because their complexity and high area/power demands are ill-suited for IoT devices. Thus, lightweight security solutions that can be easily implemented in these devices, without affecting their basic functionality, are needed.
Lightweight cryptography provides algorithms that can be harnessed by IoT-based healthcare implementations. These cryptographic primitives have simple computational needs; thus, they can be area and power efficient. Their designs on hardware are commonly utilized for IoT systems and are suitable for various health applications that focus on either low-area or high throughput. In other healthcare application cases, the trade-off between these implementational traits and the overall balance of the hardware performance is more important. A drawback to the hardware implementation of lightweight algorithms is the reduction in the security, compared to the familiar heavy cryptographic techniques, because of the decrease in computational complexity. This can be described as a hardware limitation in the IoT environment, namely, the lack of high-level security is directly intertwined with the employed resource-constrained devices' efficiency and hardware traits. Nevertheless, it does not negate the fact that these lightweight alternatives can still provide effective security and fulfil, to some extent, the requirements for a trustworthy e-healthcare system lowing an extensive investigation of the implemented schemes, proposes the best cryptographic algorithms for IoT devices.

IoT-Based Healthcare
A general architecture for an IoT-based healthcare application, as it is depicted in Figure 1, consists of IoT devices and sensors that collect health data from the environment and then transmit them to the cloud services through communication networks. These cloud services make the health data easily accessible to authorized users and healthcare providers. There are two distinct designs, the smart hospital architecture and the personalized smart health architecture, which are connected via the Internet. They both have a similar IoT-based structure. The first layer consists of various IoT devices and perhaps some hospital equipment enhanced with various sensors and actuators. It is mostly composed of health sensors, such as insulin pumps, heart rate sensors and EEG sensors; environmental monitoring sensors, such as air pressure and humidity sensors; position/motion sensors; clinical beds and diagnosis machines. All these devices are interconnected and can directly exchange data via a simple network, such as Bluetooth, ZigBee, Radio Frequency Identification (RFID) and 6LoWPAN. Furthermore, they can communicate with the external environment through the network layer of the architecture, consisting of certain networks, such as Wi-Fi and 5G, and connect to the edge/fog computing and cloud computing layers. In some cases, there is an intermediate or gateway device, namely a personal computing device, such as a mobile phone, between the first and the communication layer. The first-layer devices have very limited power and can sometimes only connect to simple, local networks. Their capabilities and level of security are not fitting for wide-area networks and the Internet. Therefore, via this gateway device, which can easily connect to more power-demanding networks, the resource-constrained IoT system can communicate with the external environment.
The edge/fog computing layer consists of edge and fog computing devices that are connected via the Internet to medical databases and cloud computing servers, namely, the last layer, where an extensive medical history of all patients is maintained and most computations are performed. Edge computing strives to resolve the problem of IoT devices' resource limitations by processing the large amount of collected data at the architecture's network edge before their transmission to the cloud. Thus, it reduces the network load and response latency [3]. In many applications, the hospital/personal computer and the personal computing devices are considered a part of the edge computing layer as they have more computational and memory resources than other IoT devices. They can easily provide solutions to the network edge by executing specific services and operations that are quite resource demanding. Nevertheless, even these devices have There are two distinct designs, the smart hospital architecture and the personalized smart health architecture, which are connected via the Internet. They both have a similar IoT-based structure. The first layer consists of various IoT devices and perhaps some hospital equipment enhanced with various sensors and actuators. It is mostly composed of health sensors, such as insulin pumps, heart rate sensors and EEG sensors; environmental monitoring sensors, such as air pressure and humidity sensors; position/motion sensors; clinical beds and diagnosis machines. All these devices are interconnected and can directly exchange data via a simple network, such as Bluetooth, ZigBee, Radio Frequency Identification (RFID) and 6LoWPAN. Furthermore, they can communicate with the external environment through the network layer of the architecture, consisting of certain networks, such as Wi-Fi and 5G, and connect to the edge/fog computing and cloud computing layers. In some cases, there is an intermediate or gateway device, namely a personal computing device, such as a mobile phone, between the first and the communication layer. The firstlayer devices have very limited power and can sometimes only connect to simple, local networks. Their capabilities and level of security are not fitting for wide-area networks and the Internet. Therefore, via this gateway device, which can easily connect to more power-demanding networks, the resource-constrained IoT system can communicate with the external environment.
The edge/fog computing layer consists of edge and fog computing devices that are connected via the Internet to medical databases and cloud computing servers, namely, the last layer, where an extensive medical history of all patients is maintained and most computations are performed. Edge computing strives to resolve the problem of IoT devices' resource limitations by processing the large amount of collected data at the architecture's network edge before their transmission to the cloud. Thus, it reduces the network load and response latency [3]. In many applications, the hospital/personal computer and the personal computing devices are considered a part of the edge computing layer as they have more computational and memory resources than other IoT devices. They can easily provide solutions to the network edge by executing specific services and operations that are quite resource demanding. Nevertheless, even these devices have some limitations to their resources and speed. Therefore, an additional layer is necessary. The fog computing layer is an extension of the cloud layer and consists of various devices, such as routers, servers and access points (APs). This layer enhances the IoT network edge by offering cloud services and additional resources closer to the computational-constrained devices.
Hence, the latency and communicational load are reduced while allowing the employment of real-time applications and better handling of the IoT's scalability problem [3].
Even though a vast number of health data is collected and utilized by IoT devices and wearables, these components display many security concerns [2]. The IoT network presents many vulnerabilities, being easily exploitable by malicious attacks. Unauthorized parties can potentially gain access and process the data for their own benefits and goals, breaching the data privacy policy that healthcare standards require [11]. Some of the most common attacks that disrupt the confidentiality of IoT systems are eavesdropping, data transmission and traffic-monitoring attacks [12]. Furthermore, the diagnosis of the patient can be distorted by data fabrication and alteration attacks; hence, preventing proper medical treatment that can lead to potentially deadly situations. There can also be cases where a function of the IoT-based healthcare system is interrupted or delayed because of Denial-of-Service (DoS) attacks, endangering the user who heavily depends on these applications. For example, smart health applications that are responsible for instantly responding to patients' health deterioration and providing in real-time first aid or alerting medical services must not be hindered by any kind of attack at any moment of operation. Moreover, due to the scalability and heterogeneity of the IoT environment, common security schemes do not provide different mechanisms in order to cover various application requirements and circumstances at any given time. Conclusively, flexible security mechanisms for IoT in healthcare, which provide and maintain basic security concepts, such as confidentiality, integrity, availability and authentication, must be implemented.
The basic rational for security schemes in IoT-based healthcare applications is explained. This rational follows the new healthcare framework, Health 4.0 [11], which presents creative methods for efficiently implementing the IoT technology by abiding to the security, performance and application requirements a healthcare infrastructure displays. First, a security scheme must utilize an efficient and minimal number of resources for proper employment in resource-constrained IoT devices. Additionally, the performance throughput must be high enough to cope with the high network load and big data transmission. Finally, the security mechanisms must be properly applied and pass formal tests provided by international standards, such as NIST, to ensure their credibility.

Basic IoT Protocols
IoT providers created an IoT protocol architecture based on the TCP/IP protocol Stack, which is divided into five layers: application, transport, network, data link and physical [13]. In each layer, different protocols can be employed depending on the application's requirements. These protocols utilize security mechanisms and particularly implement the cryptographic algorithms that will be analyzed in further sections.
There are three main lightweight protocols that operate and provide security to multiple layers of the IoT's architecture. These protocols are smaller than the Hypertext Transfer Protocol (HTTP) and more suitable for machine-to-machine communications and lowpower-bandwidth devices [13]. The first one is the Constrained Application Protocol (CoAP), which includes HTTP features while being suitable for IoT devices. It is composed of two layers, messages and request/response, with each containing a set of methods that achieve Quality of Service (QoS). The CoAP is an application layer protocol that is employed above one of the two basic transport layer protocols: User Datagram Protocol (UDP) or Transmission Control Protocol (TCP). UDP, as a connectionless protocol, is susceptible to the utilized network protocols but has a low overhead and requires minimum memory and computational resources. TCP is a connection-oriented protocol where reliable data transmission is a priority. However, there are no security mechanisms implemented in these two transport layer protocols. Thus, for transmission security, CoAP also employs the Datagram Transport Layer Security (DTLS), which is designed for datagram-based applications and prevents a variety of network attacks, such as eavesdropping and tampering.
The second lightweight protocol utilized in IoT applications is Message Queue Telemetry Transport (MQTT). It is based on the TCP protocol and is suitable for IoT architecture as it requires a minimum bandwidth and less power consumption [13]. Similar to CoAP, MQTT is susceptible to attacks without the addition of any extra security protocols. Transport Level Security (TLS) is a stream-oriented protocol that efficiently secures the TCP and allows three levels of QoS. Moreover, the employment of X.509 certificates further enhances the privacy and client authentication. Nevertheless, the utilization of long strings and even the TCP features can be difficult to implement in a very resource-constrained IoT environment. Many different variations of this protocol have been created in recent years, depending on the performance requirements of each application. Some examples are the Advanced Message Queueing Protocol (AMQP), MQTT Sensor Networks (MQTT-SN) and Secure MQTT.
The last protocol that also utilizes the TLS protocol for security is the Extensible Messaging and Presence Protocol (XMPP). This protocol follows a distributed network architecture and can be widely utilized for instant messaging functionality [14]. Overall, all application and transport layer protocols have their own advantages and drawbacks and are best suited to different IoT environments and circumstances. Therefore, always depending on the smart health application requirements, the most appropriate protocol must be selected.
Lastly, there are some additional protocols that mostly operate in the last layers of the architecture, namely, the physical, data link and network layers. The physical layer is characterized by the IoT devices that collect and transmit health data. Its security is ensured through cryptographic algorithms, which must be lightweight and require a low bandwidth and energy. The data link layer is composed of wireless communication protocols, such as Bluetooth and Wi-Fi, which provide proper security via symmetric cryptographic mechanisms. Finally, an exemplary lightweight protocol that provides security to the network layer is 6LoWPAN. It has been designed for IoT communications and offers various address lengths and a low bandwidth. The more well-known IPv6 network protocol requires more energy resources, deeming it unsuitable for low-powered IoT devices.

Architecture of Lightweight Block Ciphers
Block ciphers encrypt the data block by block using a secret key and various mathematical rounds, depending on the lightweight algorithm. The block ciphers, whose security optimizations and hardware designs are discussed, are Advanced Encryption Standard (AES), CLEFIA, KASUMI, Lightweight Encryption Algorithm (LEA), LED, Piccolo, PRESENT, RECTANGLE, SKINNY and SIMON. Finally, the designs' characteristics are collected in Table 1.
Each lightweight cryptographic algorithm has specific characteristics that can provide security for IoT-based applications. In this section, creative optimizations that enhance and verify the security were highlighted. It must be mentioned that a characteristic that reflects the magnitude of the security is the key size. Specifically, larger-sized keys are better than smaller ones.
The most utilized block cipher is the AES symmetric cryptographic algorithm. Three FPGA-based implementations employ the AES primitive with various hardware optimizations. Two of those examine the security of their design through image encryption. In addition to AES, CLEFIA algorithm's FPGA-based architecture increases its security by supporting encryption with three different key sizes. Another cryptographic primitive, which enhances the robustness of its security with a chaotic generator, is the KASUMI algorithm. The design efficiently passes various NIST tests. Additional primitives with interesting designs are the LEA and LED whose hardware implementations are straightforward. The LEA algorithm supports three different key sizes and is evaluated by image encryption, while one of the LED algorithm's designs is improved by the simultaneous utilization of two different key sizes. Lastly, PRESENT, SKINNY and SIMON implementations were analyzed. The first one presents methods for either resisting CPA attacks or detecting side-channel attacks by inserting a threshold implementation-based component. Moreover, SKINNY implements and validates a concurrent error-detection method and SIMON design enhances the security by supporting keys of different sizes, whose utilization depend on the application requirements.

Hardware Implementation Analysis of Block Ciphers
All the presented hardware designs of each block cipher displayed better hardware attributes than the conventional implementations of those cryptographic algorithms. Overall, based on the designs' results in both ASIC and FPGA, all these block ciphers can be efficiently utilized for IoT-based applications. However, some algorithms displayed better results than others. For both FPGA and ASIC structures, the algorithm with the least clock cycles and the largest key size was the KASUMI cipher. Furthermore, the algorithm that occupied the smallest area was the RECTANGLE cipher. The highest throughput was achieved by SKINNY-128 and LEA on FPGA and ASIC, respectively, and the lowest power consumption was acquired by AES on ASIC and CLEFIA on FPGA. However, not all designs were implemented in the same FPGA; thus, a comparison of them based on the FPGA board was presented. The RECTANGLE algorithm was the most area-efficient cipher on both the Virtex-5 FPGA board and ASIC CMOS-180nm. Moreover, KASUMI had the highest throughput on Virtex-5. AES-256 and LED-64 utilized the least number of slices on Virtex-7 and Artix-7, respectively. However, CLEFIA had a slightly higher throughput than LED-64 on Artix-7. The LED-64 cipher also had the smallest area and the highest throughput on Spartan-3. Finally, the PRESENT-80 had the least number of total slices and the highest throughput on Kintex-7.
Nonetheless, the efficiency of a cryptographic algorithm's implementation is defined by the performance trade-offs and not only by a single characteristic. Overall, the KASUMI cipher is the best candidate for IoT-based implementations because it displays a high throughput for a high frequency while the area and clock cycles remain low. Even when the security is enhanced and the frequency decreases, it preserves a decent throughput and a logical area utilization. In addition to KASUMI, other block ciphers are also efficient for different scenarios. CLEFIA has better throughput at a low frequency than the rest of the FPGA implementations, thus can be easily applied to IoT devices. Furthermore, the AES and SKINNY ciphers can be effectively employed in high-throughput applications that do not prioritize the area. Nevertheless, IoT is mostly characterized by resource-constrained components with energy limitations. Therefore, such approaches are not advisable. For small-area-demanding applications, the LED and PRESENT ciphers are better suited, while also retaining a high throughput. The rest of the ciphers occupy a good number of slices, but have lower throughputs, apart from LEA that displays a satisfactory throughput for a slightly larger area.

Architecture of Lightweight Stream Ciphers
Stream ciphers encrypt data by generating a pseudo-random key bit stream and combining it with the plaintext digit by digit. These primitives are also more appropriate for telecommunication standards than block ciphers because of their resistance to error propagation and efficient hardware implementations [23]. In this section, the stream ciphers whose security and designs were analyzed are A5/3, ChaCha8, Grain-v1, LIZARD, Mutual Irregular Clocking KEY (Mickey) 2.0, Rabbit, SNOW-3G, ZUC, Trivium and Welch-Gong (WG). Lastly, all the implementations are collectively displayed in Table 2. The stream ciphers are efficiently implemented on FPGA for the purpose of providing security to IoT-based applications. Few of the designs, specifically SNOW-3G, ZUC-256 and SNOW-ZUC, add extra security elements to their structures and further validate their security performance by image encryption or NIST tests. The implementation of the SNOW-3G cipher achieves an efficient security level against cryptanalysis attacks with the use of a Hyper-Chaotic Pseudo-Random-Number Generator (HC-PRNG). Thus, the architecture enhances its robustness and randomness and also passes all NIST statistical tests. Furthermore, one implementation of the ZUC-256 cipher examines the correctness of the key stream generation and validates its security. Lastly, SNOW-ZUC, which originates from a combination of SNOW-3G and ZUC stream ciphers and is suited for embedded applications, improves the randomness of the design through a chaotic generator while NIST tests verify the security.
The rest of the cryptographic algorithms utilize basic-structure optimizations, such as pipeline, serial and parallel methodologies. A unique technique is used for the implementation of the A5/3 cipher, whose architecture is based upon an optimized KASUMI cipher block. Nevertheless, all designs are verified in FPGA and offer a basic level of security to embedded applications.

Hardware Implementation Analysis of Stream Ciphers
The resulting designs of all the presented stream ciphers, similar to the block ciphers, display better characteristics than the conventional designs of the same ciphers. First, the implementations with the smallest utilized area are the serial versions of Trivium. Other relatively small-area designs are achieved by the serial versions of Grain v1. However, the Trivium cipher remains superior because it has a higher throughput than Grain v1. Second, the highest throughput is achieved by the ChaCha8 cipher, which also occupies the largest area. Thus, it is only suitable for applications that can employ larger IoT devices whose area consumption is not a priority. SNOW-3G architecture achieves a relatively high throughput for a lower frequency than the rest, classifying it as suitable for IoT devices that only operate at low frequencies. The rest of the high-throughput designs, except the parallel version of Trivium, occupy a large area without achieving a good trade-off. The parallel version of Trivium achieves the best ratio between throughput and area. It has the second-highest throughput value, medium area, an effective 80 bit-size key and an adequate frequency. The parallel version of Grain v1 and the basic version of Trivium also have efficient trade-offs between area and throughput. Nonetheless, the parallel version of Trivium is recommended if the area requirements are slightly flexible.

Designs of Hash Functions, MACs and Authenticated Schemes
Hash functions, message authentication codes and authenticated encryption schemes are cryptographic methodologies that can provide security to IoT-based applications. The first primitives, hash functions, offer the ability to compress to a specific length the transmitted data. Lesamnta-LW, LHash-96, SPONGENT-88, PHOTON-80/20/16 and sLiSCP are the hash functions whose hardware implementations are examined in the following section with all of their characteristics accumulated in Table 3. The second discussed primitives are MACs that are mainly utilized for the prevention of identity theft or message forgery between two devices. Two MACs designs, Chaskey and LightMAC, were analyzed and presented in Table 4. Finally, the designs of authenticated encryption schemes, ACORN, AEGIS, Ascorn, NORX and KETJE, which enhance the confidentiality, integrity and authenticity of the system, are displayed in Table 5.
The structures of these algorithms follow the same primary optimization techniques as block and stream ciphers, and additional security components, such as chaotic generators, were not utilized. Therefore, further analysis will not be needed. Nevertheless, all the designs are deemed suitable for providing basic security in constrained lightweight applications due to the limited resource utilization they achieve. It must also be mentioned that, usually, these security schemes cannot provide complete security to the systems alone, but they are employed together with other security primitives to properly prevent more attacks and susceptibilities.

Hardware Implementation Analysis of Hash Functions
The presented hash functions are efficiently implemented in both FPGA and ASIC. Specifically, the architecture of Spongent-88 has the smallest area and is more suited for constrained-area applications. However, it is the slowest design and has the lowest throughput, even at a high frequency. The highest throughput is achieved by the lowfrequency PHOTON-80 implementation, without occupying many slices. Furthermore, the other design of PHOTON-80/20/16 also has good throughput with a smaller number of slices, average frequency, efficient power consumption and fewer clock cycles. Overall, both PHOTO-80 schemes have greater trade-offs than the rest of the FPGA implementations, and even though they do not have the smallest area of all hash functions, they are compact enough to be employed in area-constrained IoT-based applications.

Hardware Implementation Analysis of MACs and Authenticated Encryption Schemes
Recently implemented MACs are Chaskey and LightMAC. The devices that implement these MACs are the Arduino and the NUCLEO-F401RE board, which vary from the platforms employed for the previous algorithms. Thus, the design traits are slightly different. Between these two algorithms, LightMAC excels at power consumption and throughput, while also achieving a low execution time for an average-sized message. Nevertheless, the Chaskey algorithm occupies less memory space than LightMAC and has a low execution time in Arduino. Overall, LightMAC has a more balanced implementation with efficient features.
In contrast to MACs, authenticated encryption schemes are implemented in ASIC and FPGA boards. The highest throughput, but also the largest area, is achieved by the AEGIS-128L design. Therefore, it can only be utilized in high-throughput applications where area is not a priority. The second-highest throughput and the smallest area in FPGA is achieved by the NORX design. Furthermore, the ASIC implementation of NORX has a high throughput, but a relatively large area. Some of the ASIC designs of KETJE have smaller area than NORX algorithms; nevertheless, the latter display better trade-offs and have more efficient designs in both FPGA and ASIC. The only issue is the power consumption. ACORN and KETJE have the lowest power-consumption values in FPGA and ASIC, respectively; however, the NORX scheme surpasses them in other sectors. Therefore, each scheme can be employed in an IoT healthcare system depending on the available resources and the priorities the application demands.

Comparison Results
In this section, all the lightweight cryptographic primitives are compared. Out of all the categories, stream ciphers and mainly block ciphers provided the most efficient security, which in some cases was even enhanced. As for the hardware limitations, the ASIC implementations that displayed better trade-offs were LED and LEA designs. The former is qualified for small-area applications with good throughput and power consumption, and the latter is appropriate for high-throughput applications with an average area and efficient power consumption. The Piccolo-80 and KASUMI implementations on Virtex-5 are the most suitable for area-efficient and high-throughput applications, respectively, without lacking in other sectors. For Virtex-7, the NORX design is the most beneficial. The other designs in this board demonstrate high-throughput results, but also large areas, which are unacceptable in the IoT resource-constrained environment. Furthermore, the LED-64 design had the best trade-offs on the Artix-7 board, while the PRESENT cipher was deemed the most balanced architecture on the Kintex-7 board. PHOTON-80 function on Spartan-3, ZUC on Spartan-6 and Trivium on Spartan-7 exhibited good throughput and overall better characteristics than the rest of the ciphers on the same FPGA boards. Lastly, between the two MACs, LightMAC had better trade-offs than Chaskey.
After extensive research, the most efficient algorithms out of all those presented in this paper were the KASUMI block cipher, PHOTON-80 hash function, PRESENT block cipher and Trivium stream cipher. The most rapid implementation with the most enhanced security was the KASUMI block cipher. It has a larger area than most but compensates in throughput capability. Moreover, the design with the chaotic generator operated efficiently at a low frequency, which is common in IoT. Nevertheless, the other designs had better performance trade-offs. The PHOTON-80 hash function displayed even higher throughput and smaller area for an equally low frequency, deeming it suitable for low-area and highthroughput-demanding applications. An even smaller area was obtained by the structure of the Trivium stream cipher. This architecture has the highest throughput of all with an average frequency. Overall, Trivium had the best ratio between area and throughput. Finally, the smallest area of these four algorithms with a relatively high throughput was achieved by the PRESENT block cipher. The throughput may be the lowest of these four but compared to algorithms with the same number of total slices it was the highest.

Conclusions and Outlook
IoT-based healthcare applications have a variety of vulnerabilities and susceptibilities that can potentially endanger the user's wellbeing due to the considerable impact these health services have on the patient's everyday life and the large amount of private health data collected. Many security mechanisms that depend on lightweight cryptographic primitives have been introduced to the literature. These implemented security schemes aim at ensuring the privacy, confidentiality and integrity of the system, while also utilizing few hardware and energy resources and achieving high-throughput and performance efficiencies. Nevertheless, depending on the application's performance, employed resources and security requirements, specific cryptographic algorithms display better trade-offs between hardware efficiency, performance throughput and security. Therefore, in this paper, after extensive research and the thorough analysis of all recent implemented lightweight cryptographic primitives, four algorithms were deemed better suited for IoT devices in healthcare applications. These algorithms were the KASUMI block cipher, PRESENT block cipher, Trivium stream cipher and PHOTON-80 hash function.
Overall, this paper presented an analysis of the IoT-based healthcare design and the research conducted to date on hardware implementations of cryptographic algorithms. Specifically, the capabilities, to date, and therefore limitations of IoT-based health applications based on the hardware designs of lightweight cryptographic primitives were demonstrated. For future directions, these conclusions will be considered with the aspirations of improving and simplifying IoT-based implementations for security purposes in the healthcare domain.
Author Contributions: K.T. and N.S. contributed to the research, investigation, results analysis, resources and writing-review and editing. All authors have read and agreed to the published version of the manuscript.