Dimensions of artificial intelligence techniques, blockchain, and cyber security in the Internet of medical things: Opportunities, challenges, and future directions

1 Introduction

The Internet of medical things (IoMT) brings together the Internet of things (IoT) and medical equipment. The IoMT is expected to form the foundation of future healthcare systems in which all medical equipment are connected to the Internet and operating under the supervision of medical experts. As it develops, the IoMT offers quick and less-expensive healthcare [1]. IoMT-enabled devices include smart watches, shoes, electroencephalogram and electrocardiography (ECG) machines, as well as airflow, blood pressure, and motion sensors [2]. Through the Internet, the vital signs collected from the sensors are sent to the IoMT application, which transmits the data to healthcare professionals and medical staff and sends a response back to the patients who need the information. Sensor data are transmitted straight to a large cloud storage space. The design of the IoMT is made up of three layers: the things, the fog, and the clouds. Patient monitoring devices, sensors, medical records, and others are in the things layer, which is in direct contact with the users of the ecosystem. This layer is where the data from the sensors is collected. The devices used for it should be in a safe place to ensure the accuracy of the data collected. The fog layer, which is between the cloud and things layers, is composed of servers and gateway nodes for a fog networking framework with few nodes. The cloud layer is made up of places to store data and ways to process it so that it can be analyzed and used to make decisions [1]. As IoMT devices are vulnerable to cyberattacks, privacy and security are problems and challenges faced by healthcare providers and patients. A safe IoMT environment has been created by combining blockchain (BC) technology with artificial intelligence (AI).

The IoMT uses BC technology to address security issues and protect the privacy of its users. BC is mainly used to ensure that digital money transfers are safe. It is a digital log that is open and distributed to the public and is used by a peer-to-peer network to store data. Block addressing is done with public key cryptography to ensure that everyone can see the data at the block level. Thus, BC gets rid of the risks associated with centralizing data, such as manipulation, and it is a good idea to set up an IoMT that is safe and protects privacy [3]. AI is a promising technology that will change the IoMT.

Machine learning (ML) is a type of AI that enables machines to learn from their experiences to improve their performance in the future without any help from humans [4]. ML is considered the main technology of modern network administration. Many IoMT systems are becoming complex, dynamic, and heterogeneous, and consequently, they are difficult to administer. To attract consumers, these IoMT services must also be improved in terms of efficiency and diversity.

Thus, we may conclude that the IoMT can benefit from the support provided by ML. Using ML in IoMT enables users to gather deep analytics and develop intelligent IoT apps that are both efficient and effective because ML can facilitate the extraction of information and characteristics hidden in IoMT data [5]. ML algorithms and AI have opened new avenues in IoMT. ML makes it possible to learn from data and find meaningful patterns because IoMT device data are recorded in a database and are easy to use in prediction. The patient’s medical information and outcomes assist AI and ML systems in forecasting better treatment choices, thereby assisting in effective monitoring and recovery. AI-controlled emergency traffic helps ambulances and other emergency services. AI and deep learning (DL) are innovative ways to investigate “normal” and “strange” behavior. Each aspect of the IoMT framework may be collected and analyzed to determine typical examples of communication and identify harmful behavior at the proper moment. ML and DL strategies can also help predict future attacks, which are often transformations of previous ones, by learning from current models. IoMT frameworks must evolve beyond supporting safe communication between devices to the knowledge based on the security enabled by ML/DL methods for frameworks that are applicable and safe [6,7].

In this study, we present a systematic review of the most recent and important studies and research methods in the IoMT. Healthcare facilities and patients face privacy and security issues in IoMT networks because IoMT devices are vulnerable to cyberattacks. The study aims to discuss the role of AI, BC, and cybersecurity technologies, and discuss strategies for integrating these technologies, as well as the challenges, opportunities, and future directions of research on the IoMT. Numerous IoMT-related research questions exist such as the following: (i) What are the basics of AI, BC, and cybersecurity and how can we benefit from them? (ii) What is the effect of AI, BC, and cybersecurity in the IoMT? (iii) How can the integration of AI, BC, and cybersecurity technologies help in developing the IoMT? (iv) What are the main challenges, opportunities, and future directions for the IoMT? (v) What are the obstacles to the use of AI, BC, and cybersecurity and how can they be overcome?

The main contributions of this study are summarized as follows:

• This study analyzes the effect of AI-IoMT, BC-IoMT, and cybersecurity-IoMT technologies.

• This review discusses strategies for integrating AI, BC, and cybersecurity technologies that can aid in the development of emerging systems employing a decentralized architecture especially in healthcare applications.

• This study provides an overview of AI, BC, and cybersecurity basics, their main characteristics, and how they could be utilized for IoMT.

• Our study contributes to the removal of obstacles to the use and application of AI, BC, and cybersecurity in healthcare applications.

• This study highlights the research challenges and concerns associated with the introduction and use of AI, BC, and cybersecurity in healthcare applications.

The rest of this study is organized as follows. Section 2 presents the role of AI techniques in the IoMT. Section 3 shows the impact of blockchain in the IoMT. Section 4 describes the role of cybersecurity in the IoMT. Section 5 presents the opportunities, and Section 6 sums up the challenges in the IoMT. Section 7 presents the future research directions. We suggest novel techniques for IoMT in Section 8. Section 9 discusses the results of the study, and Section 10 presents the conclusion.

2 Role of AI techniques in IoMT

AI is a modern computer science technique that develops software and methods that make machines smart and effective in tasks that typically require expert human intelligence. DL, ML, and traditional neural networks are AI subsets with distinct skills and functionalities that have the potential to enhance the effectiveness of advanced medical sciences. These sophisticated systems make clinical diagnostics, medical imaging, and human decision-making easier. At the same time, IoMT integrates software applications and network-connected biomedical devices to improve human health. AI-based models have proven their worth in the healthcare and pharmaceutical sectors by improving the real-time health monitoring, efficiency of therapeutic drug manufacturing, and predictive forecasting [8]. Both AI and the IoMT face certain challenges, which become more complex when these technologies are combined. Among these challenges are artificial stupidity, security, compatibility and complexity, lack of confidence, and cloud attacks. Artificial stupidity refers to a computer program’s failure to complete base tasks perfectly. To analyze and understand data and make highly accurate and sane decisions, AI systems and the algorithms they use should be well developed. Because AI and the IoMT gather sensitive and important data from their clients or users, the data have to be in secure and safe hands. However, we do not know when attackers can access sensitive and important data, which is why security is the most important issue with any technology. The IoMT represents the fusion of several devices with various technologies, which may result in a variety of issues when all these devices are combined into one system.

More complex ecosystems also exist due to the proliferation of various devices. Because IoMT is a new and emerging technology, both consumers and businesses are concerned about security and lack confidence in the ability to secure IoMT devices and the integrity of the information generated [9]. According to Lakhan et al. [10], the proposed bio-inspired robotic enabled in BC-fog-cloud-assisted IoMT has been used to collect healthcare data. The IoMT consists of sensors on the fog-cloud computing nodes with BC enabled to minimize the execution costs and BC of healthcare uses. A new study by Lakhan et al. [11] suggested a novel deep reinforcement learning (DRL) and BC-enabled health care system in the IoMT, which includes offloading based on DRL policies and BC task scheduling. Also, a new study by Samuel et al. [12] suggested a federated learning (FL)-BC-based privacy and information security infrastructure to resolve the issue of big data repositories while protecting the owner’s data privacy. Rahmadika [13] used lightweight sensor nodes and FL- and BC-based privacy preservation strategies to secure wireless network-based malware detection. Saheed and Arowolo [14] presented a classifying model based on supervised ML models and deep recurrent neural network (DRNN) for classifying and forecasting unexpected cyberattacks.

Nayak et al. [15] developed a new hybrid optimization and lightweight extreme learning machine (ELM)-based framework to detect unauthorized entry into the ecosystem. Also, a new study by Liu and Li [16] presents a BC- and DRL-enabled IoMT system. To process the security problems, the BC method is used, and a DRL is applied to optimize the system power performance and efficiency and security using both BC technology and big data analytics powered by AI. A private BC-based protection system was proposed for home surveillance in an IoMT study [17]. Moreover, Kumar et al. [18] proposed novel methods for ameliorating security in IoMT. An improved Elman neural network algorithm was used to analyze how sensitive the data is for using BC technology in data storage. Lakhan et al. [19] devised a task-scheduling framework to discover and guarantee data privacy and fraud on fog-cloud nodes through an FL-based BC-enabled task-scheduling. Lakhan et al. [20] developed a serverless system and Boltzmann machine-based scheduling algorithms that minimize execution time and cost while still meeting timeline, cost, and security requirements. Sanaat and Zaidi [21] suggested a cloud-based ongoing learning strategy to aggregate full-dose from low-dose positron emission tomography images. Lin et al. [22] proposed a task offloading and resource allocation mechanism based on BC technology to improve system security using collective reinforcement learning (CRL). Su et al. [23] suggested a novel technique that uses the relationship between human blood flow and body surface temperature to quickly detect heart valve diseases by employing a DL algorithm to improve curve fitting and analysis. A DL-based encryption-decryption network is proposed by Ding et al. [24] to transfer from their original domain-to-target domain securely.

Rahman and Shamim Hossain [25] presented an edge-IoMT system that employs DL to predict health problems related to COVID-19 and generate reporting and notifications that can be used to help doctors make decisions. Alqaralleh et al. [26] introduced an effective model based on a BC-assisted model for diagnosis and secure image transfer in the IoMT environment. In patient monitoring such as that reported by Ahmed et al. [27], a DL-based algorithm was used to present an IoMT-based framework for nonsurgical automated patient discomfort monitoring detection. Another study by Mahanty et al. [28] proposed an IoMT-based system that blends deep-transfer learning (DTL) model with a fuzzy ensemble technique for classifying chest CT scans. Also, Akhtar et al. [29] proposed an IoMT-based framework for health care monitoring with improved recurrent neural network (I-RNN)-based classification to predict multiple diseases. Moqurrab et al. [30] proposed a new fog-enabled privacy-preserving framework called adaptive-neuro fuzzy classifier to optimize the health system through DL. Yacin Sikkandar et al. [31] combined the GrabCut method and adaptive-neuro fuzzy classifier to create a new segmentation-based classification method for skin lesion diagnostics. In the cancer diagnosis presented by Khamparia et al. [32], a DL IoMT-based approach for detecting and classifying cervical cancer cells was developed. A summary of studies using AI in IoMT is shown in Table 1.

Referencing Table 1, we can conclude that several studies [14,15,20,21,23,24,25,26,27,28,29,30,31,32] have found that ML and DL are good at making accurate predictions, but they have problems such as having a low level of security and not taking cost or time into account. The remaining studies [10,11,12,13,16,17,18,19,22,23,24,25,26] using ML and DL are based on BC outperforming other methods in terms of security but suffering from high cost and latency.

3 I Impact of blockchain on IoMT

BC is an innovative technology that is used to produce innovative solutions in numerous industries, including healthcare. A BC network is applied in the healthcare system to share and preserve patient data across doctors, hospitals, pharmacies, and laboratories. BC applications can precisely detect grave errors, including potentially fatal ones in the field of medicine. Therefore, it can improve the effectiveness, safety, and openness of sharing health data in the healthcare system. By using this technology, medical institutions can improve their analysis of patient data [33]. The limitations of the BC system in the IoMT are scalability and rising overhead or computational resources. Scalability is an issue in BC systems because the number of users raises the computational requirements of the infrastructure. When a large number of smart equipment-sensors exist, their processing capacity is less than that of a typical computer, offloading a significant portion of the resource demands to other computers, like an end-device or the cloud.

In general, numerous BC solutions are computationally expensive and necessitate a substantial bandwidth average, resulting in data delays and substantial processing power. Such requirements are infeasible for the vast majority of IoMT devices, which are primarily sensors with low processing capability. This decrease in processing speed may lead devices to function suboptimally or even overburden the device to the point that it cannot operate the source code or BC program [34]. A cost-effective BC-enabled scheduling method was proposed by Lakhan et al. [35] to minimize execution cost and latency. To guarantee data security and node-to-node validations throughout the system, the studies [36,37] developed a BC-enabled socket that is decentralized, secure, cost-effective, and powerful for health care applications. Lakhan et al. [38] used BC-enabled smart contracts and a cost-effective scheduling scheme to develop an IoMT system. With symmetric key cryptography, consistency of data and validation is guaranteed. Lakhan et al. [39] introduced a function-based task scheduling framework that uses BC technology to choose services with efficient cost and scheduling tasks based on the cost of execution and quality of service. Arul et al. [40] proposed “a multi-modal secure data dissemination framework” relying on BC in IoMT to control and obtain secure patient records to protect the privacy of each person’s health transactions before they are shared. Ayub Khan et al. [41] suggested using a BC with the NuCypher re-encryption process to boost security and offer a medical ledger that is transparent and truthful.

With regard to security issues, Garg et al. [42] proposed a consortium structure based on BC hyperledger fabric that provides integrity, security, and transparency in a secure serverless environment by designing an authentication key agreement protocol. Wu et al. [43] presented a framework BC compatible to publishing data and sharing it in IoMT, with the goal of providing security and privacy protection for various apps. Another proposed study by Wang et al. [44] was a BC-based sharing system through IoMT, which combines cryptography basics with interplanetary file system (IPFS) to keep IoMT data safe. Kumar and Tripathi [45] developed a novel distributed structure that relies on smart contract IPFS cluster nodes for use in healthcare system authentication and controls access to achieve the security requirement of IoMT. By using the advanced-encryption standard (AES) encryption mechanism integrated into mobile devices (MD) to secure patient information, Nguyen et al. [46] proposed a novel distributed health architecture that merges mobile-edge computing and BC to protect data offloading and sharing in remote hospital networks. Ismail et al. [47] suggested a lightweight BC design that reduced the computation cost for the management of health care data. This architecture offers privacy and security by analyzing various attacks and threats. Another study used BC with cloud [48] to address BC-based cloud-centric IoMT health care systems’ larger memory costs, latency, and single point of failure. Another study proposed the hybrid computing paradigm with a BC-based distributed data storage system to develop a health data privacy strategy and medical devices. Malamas et al. [49] proposed a framework for managing both IoMT devices and health records using blockchain.

By using BC technology, Dilawar et al. [50] proposed a secure mechanism for providing authenticity, confidentiality, and integrity of data transfer in IoMT. To enable secure electronic health record (EHR) sharing across various health care providers and patients, Nguyen et al. [51] proposed a new method for sharing EHRs that combines IPFS and BC on a mobile cloud service. Another study by Babu et al. [52] suggested a BC-based permissioned framework for securely sharing information that offers high certainty and ensures the provenance integrity of data. Also, Li [53] suggested a BC-based medical prescription system that can provide online prescription processing, access rights, and identity verification to secure patients’ privacy. To maintain the security of data transit and information management among nodes that are interconnected, as well as to improve information accessibility and the scalability of the health care environment, the BC-assisted secure information control framework has been proposed [54]. The purpose of using BC in the work of Sultana et al. [55] was to combine BC technology with zero-trust principles to improve transfer images and the security of health information. The studies summarized using BC are shown in Table 2.

From Table 2, we conclude that most studies use the HB/ECG datasets and outperform in terms of security and privacy, and [39,48,53] outperform in accuracy. However, [35,36,38,42,44,47,50,51,54] suffer from high energy and resource consumption, while [37,41,45,53] suffer from high cost and [40,49] from storage concerns. The rest of the studies have communication and cost issues.

4 Role of cybersecurity in IoMT

The IoMT is a subset of the IoT in which medical devices exchange confidential data. With these improvements, the healthcare industry can provide better care to its patients. Security is seen as a major problem for any technology that depends on the IoT. Attackers can do many different things that can cause security problems. Attacks such as remote hijacking, impersonation, denial of service (DOS), predicting passwords, and man in the middle all pose security risks. In the case of such attacks, sensitive data connected to the IoT could be leaked, changed, or even made inaccessible to authorized users [14].

Many cybersecurity challenges are involved in the IoMT. First is the number of vulnerable IoMT devices. Because of the lack of processing, storage, and energy resources in providing a higher level of security, such flaws could result in extensive security breaches and substantial financial losses. Second, massive numbers of IoMT devices that are Internet connected can rapidly generate vast quantities of data. Because it is impractical to store all data in centralized locations for processing, as is the case with existing cloud computing systems, these data must be processed and stored in several decentralized edge clouds or edge-computing nodes. How to improve data security and privacy in a distributed-edge computing environment is still an unaddressed and difficult issue. Finally, it is common for IoMT devices with limited resources to send tasks to an edge-computing platform with many resources so that they can be done quickly. Collaboration and offloading of tasks may raise more security issues [56]. For instance, Lakhan et al. [57] suggested a system to execute workflow apps on extra nodes with the least amount of system security risk and delay when offloading and scheduling. Lakhan et al. [58] presented a secure offloading system that uses network monitoring and local device surfing profile to identify and foresee any attack before offloading. This study aims to reduce node energy consumption and move data securely between nodes throughout the application partitioning inside the system [59].

Chhabra and Lata [60] suggested obfuscation-authentication-based AES to keep medical imaging systems in IoMT safe and to offer better resistance to known assaults and successfully secure health records before sending. Furthermore, Karmakar et al. [61] devised a security infrastructure for IoMT that is confidentiality preserving, secure, and trusted. It employed a mechanism called elliptic curve cryptography (ECC) to verify medical equipment and communications. Masud et al. [62] designed a lightweight security protocol for establishing masks between a sensor node and a doctor. The summary of studies using cybersecurity in IoMT is shown in Table 3.

Based on Table 3, no study has proposed AI techniques such as DL or ML to enable cybersecurity. Instead, research [59,60,61,62] used lightweight algorithms that outperformed traditional algorithms in terms of reduced energy and resource consumption. However, they were more expensive and suffered from security problems, while other algorithms [57,58] outperformed them in processing time.

5 Cybersecurity opportunities in IoMT

Opportunities for new research and innovation in cybersecurity, IoMT, and edge-computing are anticipated as a result of the development of BC technology, AI, and ML [56]. This section presents a summary of new opportunities in this field.

6 Cybersecurity challenges in IoMT

This section discusses some of the anticipated problems involved in harmonizing and incorporating AI, BC, and cybersecurity to create IoMT infrastructure for healthcare applications. AI can run on BC and use the network’s shared data to generate predictions. BC is a cybersecurity system that employs node-to-node bonds to link old blocks to new ones in order. Healthcare, cybersecurity, predictions, cryptocurrency, and other companies benefit tech convergence. The more these industries employ digital technology to offer services, the larger the risk of hacking. Despite their openness and integrity, BC systems encounter cybersecurity and data-security risks. The decentralized access strategy to public BC has been misused for criminal behavior, notably with cryptocurrencies [64].

The following are some AI, BC, and cybersecurity challenges:

• Scalability

  Scalability and increasing processing resources in an IoMT setting are some BC restrictions. Scalability is a challenge in BC systems because the number of users raises the computational needs of the infrastructure. The situation becomes more difficult if many smart devices or sensors are connected because their processing capabilities are smaller than those of a traditional computer, and much of the resource requirements have to be uploaded to an edge or cloud device [34].

• Confidentiality

  Acquired data are available to all consumers on public BC ledgers, which makes it possible to process the data in safely and legally. Secure BC systems make it hard for AI to access and share the huge amount of data it might need to make decisions [64].

• Integrity

  Even when privacy is assured, data are not always safe from outside interference such as adding or changing some fragments to change the data. When it comes to life-critical applications, such attacks are dangerous. Data loss can happen when a communication network is not sufficiently secure [65].

• Availability

  In IoMT applications such as emergency treatment services and real-time monitoring, the patient’s data should be accessible around the clock, 7 days a week, so that doctors and physicians can take appropriate action. The lack of node or network availability at the perception or processing layer causes the inaccessibility of vital data, which may result in a patient’s death [66].

• Privacy

  IoMT devices used in healthcare communicate with one another through broadcasting, which may cause problems and threats to privacy in the long run. A person who uses a wireless device could face serious problems such as having their location tracked. An adversary can eavesdrop on a conversation and have access to vital information and cause significant harm to patients. When accessing data, IoMT-based health systems must adhere to privacy regulations and policies. A data access request should be compared to policies that have already been set up to decide whether to grant access. This framework makes it hard to use the IoMT architecture in healthcare applications. First, data have to be gathered and processed safely in real time. Also, when different domains work together, finding various types of data sources is important. It is also necessary to break some privacy regulations to meet performance standards. For example, reporting heart arrhythmia in real time is more important than meeting privacy requirements. Thus, it is important to let some privacy breaches send an urgent message to healthcare providers. IoMT devices are vulnerable to privacy attacks that can gather information by listening to encrypted radio transmissions from sensors. This type of attack is called fingerprint and timing-based snooping. All that is needed to start it is the time each message was sent and its fingerprint. However, there are several ways to defend against this kind of attack. It is possible to reduce the signal strength outside the building to make it more likely that an eavesdropper will lose packets. Also, the sensor can send radio messages even when it does not have any new information to send. Another possible solution is to randomly delay messages so that they are hidden when an event happens. It is even possible to hide the transmitter’s unique mark [65].

• Security

  Dealing with cyberthreats is another IoMT security issue that has not been solved yet. Security attacks are difficult to stop because sensor devices have limited resources, they are hard to reach, and wireless communication is not always reliable. Attackers can also take advantage of the fact that IoMT-based systems often fail randomly. For example, an attacker could make up or alter data or launch a DoS attack. Thus, a patient could get the wrong treatment or no treatment at all [65].

• Latency

  The network latency is a crucial part of how well the healthcare application works. Latency has different effects on various types of healthcare applications. It has a strong effect on emergency services and multimedia healthcare applications such as live surgeries [66].

• Energy efficiency

  Batteries are used to provide power to the sensors. If the battery dies, the sensor nodes will stop functioning, which will be detrimental to the ability of the system to assist people in times of medical crisis [65].

• Absence of compliance, standardization, and protocols

  There are no universally accepted standards for BC applications to date. Connectivity, administration, interactivity, and architectural standards for blockchains are now under development by international telecommunication union, institute of electrical and electronics engineers, national institute of standards and technology, and many other standard-setting organizations. Public-BC financial transaction funding and automated fund transmission using cryptocurrencies require both local and global governments and organizational regulations, rules, laws, and procedures for BC implementation, arbitration, and conflict resolution in the context of AI applications [64].

• Quantum computing

  Quantum computing involves deciphering public-key encryption to reveal hidden keys. The BC infrastructure currently uses encrypted digital signatures. Insiders believe that by 2027, quantum computing will guarantee that the security of BC is impossible to break [64].

7 Future directions of cybersecurity in IoMT

In this section, we discuss the potential directions of cybersecurity.

8 Suggestions

In this section, we suggest some methods to improve cybersecurity in the IoMT.

9 Discussion

The aim of this literature survey is to summarize AI role, blockchain, and cybersecurity in IoMT, and identify the strengths and weaknesses of each, as well as the dataset used. Reviewing the role of AI in the IoMT shows that most studies have used DL and ML. These studies are good at making accurate predictions, but they have problems with security and privacy and do not take cost or time into account. Regarding the role of the blockchain in IoMT, most studies have used the HB/ECG datasets, which outperform in terms of security and privacy but consume more energy and resources. For cybersecurity, most of the studies have used lightweight algorithms that reduce energy and resource consumption but are more expensive. No study has proposed AI techniques such as DL or ML to enable cybersecurity. AI, BC, and cybersecurity techniques face challenges in making IoMT infrastructure for healthcare applications, such as security, privacy, latency, energy, availability, integrity, and scalability. Security attacks are difficult to stop because sensors have limited resources and are difficult to reach, and the connection is not always reliable. In terms of privacy, IoMT devices used in healthcare communicate with each other by broadcasting. In the long run, this situation can lead to problems and privacy risks. A review of opportunities and future directions in the IoMT showed that most studies expect new possibilities for improving cybersecurity in the IoMT because of AI and BC. Despite advancements in DL-AI, trustworthiness concerns occur, especially regarding the security and privacy of AI models. Accordingly, AI models could be hijacked. BC systems can authenticate or authorize IoMT cloud data access. Using a decentralized BC system is another area where authentication or authorization could be applied. BC and AI could solve this problem.

10 Conclusion

As most IoMT devices are small and easy to access, this study concludes that the main issue is security of healthcare data, so the IoMT needs new measures to keep the system safe. The most important modern methods and technologies for overcoming network device have limitations in making the IoMT network more secure. Furthermore, the IoMT faces major challenges in security, privacy, availability, latency, and energy use. This review describes the integration schemes of AI, BC, and cybersecurity technologies, which can support the development of new systems based on a decentralized approach especially in healthcare applications. This study also identifies the strengths and weaknesses of these technologies, as well as the datasets they use. Because of the characteristics of the IoMT, future research must focus on DL/ML and BC techniques to improve cybersecurity by utilizing lightweight solutions.