A Critical Analysis on the Security Architectures of Internet of Things: The Road Ahead

1.1 Security Requirement

In IoT areas such as remote monitoring and intelligent transport systems, terminal nodes will gather information, which is then transmitted to the IoT platform, and then the platform will send commands to the terminal nodes through the relay devices. Therefore, it is difficult to protect the validity of both the terminal nodes and the platform in such networks [18]. In the security of IoT, several tasks that are to be performed are (i) inserting keying material in the development stage of the device, (ii) demanding new keying material while in operation, (iii) establishing access control policies in order to access the networks and services, (iv) using hardware security modules for protecting keys from tampering, (v) managing software update, and (vi) developing and selecting proficient cryptographic primitives. Traditional security systems provided by the IoT research community presents mainly point solutions; however, this seldom aids in understanding the big picture for securing IoT devices [20]. For the security of IoT, there are several works that are mainly focused on the security architectures and recommended countermeasures, secure communication and networking mechanisms, cryptography algorithms [52, 55], and application security solutions [27, 57].

Recent researches mainly focus on three security aspects, such as system security, network security, and application security. System security concentrates on the entire IoT system to recognize distinctive privacy and security challenges, to devise complete security frameworks, and to afford security measures and strategies. Network security mostly concentrates on wireless communication networks to frame key distribution algorithms, authentication protocols, advanced signature algorithms, access control mechanisms, and secure routing protocols. Specifically, authentication protocols are more common in dealing with security and privacy issues in IoT. Furthermore, heterogeneity and hierarchy should be considered while designing it. The major security goals of IoT are to ensure proper identity authentication mechanisms and provide confidentiality about the data, etc. The available models should implement security by making use of different requirements like data confidentiality, data integrity, data availability confidentiality, policy enforcement, and reputation [4, 30, 39, 44, 45, 46].

2.1 Lightweight-Based Security Techniques

This section deals with the security techniques related to being lightweight. There are several works that focused on authentication as a security issue in developing security protocols. Accordingly, Hernandez-Ramos et al. [16] have introduced a lightweight authentication and authorization system to maintain smart objects. Moreover, such systems were structured in a security framework that is submissive with the Architectural Reference Model (ARM) newly introduced by the EU FP7 IoT-A project. Also, Kim and Kim [21] have presented a lightweight encryption scheme called PRINCE, and a technique to enhance the message space of PRINCE by enhancing XLS (eXtension by Latin Square). They examined the cryptographic requirements of the inverse-free lightweight encryption scheme and recommended how to expand a PRINCE-like encryption scheme to secure the IoT system.

Another significant lightweight architecture to handle authentication issues is given in Ref. [8], where Fan et al. introduced a lightweight RFID mutual authentication protocol (LRMAPC) with cache in the reader. It stores the recently visited key of tags in LRMAPC, such that recently visited tags are directly authenticated in the reader. LRMAPC can greatly minimize the cost of computation and transmission. Particularly, it can largely minimize the cost of computation if there are more tags to be authenticated. Using GNY Logic, the accuracy of LRMAPC is verified. Further, Raza et al. [37] presented a lightweight 6LoWPAN compressed with IKEv2 for key management for IEEE 802.15.4 link layer security. Hou et al. [17] formulated the lightweight authentication scheme named HB-MAP. In this scheme, the client needs to store only a little amount of messages and do a few bit-wise operations. The server will do the computation-consuming operations such as the generation of random challenges. Last, they carried out security vulnerabilities analysis of the HB-MAP protocol.

Ning et al. [32] concentrated on unit IoT and ubiquitous IoT (U2IoT) architecture to devise an aggregated-proof-based hierarchical authentication scheme (APHA) for the layered networks. Actually, the aggregated proofs were launched for multiple targets to attain forward and backward unspecified data transmission. The directed path descriptors, homomorphism functions, and Chebyshev chaotic maps were employed together for mutual authentication. Several access authorities were allocated to attain hierarchical access control. In the meantime, the BAN logic formal analysis was carried out to confirm that APHA has no observable security deficiencies, and it was most likely existing for the U2IoT architecture and other IoT applications.

To handle the security and privacy issues in IoT, a lightweight no-pairing ABE scheme based on elliptic curve cryptography (ECC) was given by Yao et al. [54]. Instead of the bilinear Diffie-Hellman assumption, the security of the approach depends on the elliptic curve decisional Diffie-Hellman assumption, and was verified in the attribute-dependent selective-set model. Shi et al. [43] have presented a lightweight authentication scheme to satisfy the need of secure authentication for wireless Internet applications. The scheme was utilized in several authentication circumstances in mobile wireless Internet applications, and offered communication sessions with privacy and security from other networks. It was installed as an add-on module at the application layer, and does not require any modifications to the present Internet applications.

Lightweight security techniques typically provide the same or enhanced services as their heavier counterpart, but have a lighter footprint in various ways. Lightweight protocol codes perform faster than standard protocols. They tend to have a smaller overall size, to leave out unessential data, and might use a data compression technique to have a lighter effect on network communication.

2.2 Robust Security Techniques

There are several studies that considered authentication as an important issue and proved the security issues with a number of attacks. The attacks proved that security techniques are robust techniques in IoT. There are only a few articles in the literature dealing with robust architectures. Three recent articles on IoT are reviewed here. In view of that, Panwar and Kumar [33] have presented an effective Datagram Transport Layer Security (DTLS) mechanism that employs public certificates for authentication, which makes a robust DTLS security. They make use of a certificate authority that can provide the digital certificates to the client as well as the server, and can improve the communication efficiency. They have established a certificate of authority for preshared key communication.

Jan et al. [19] presented a robust mutual authentication approach to authenticate the identities of the contributing devices before involving them in communication. The connection overhead is minimum in this method, and it provides a robust defense solution to combat several kinds of attacks. Leo et al. [26] developed a combined architecture for data and service transactions in IoT scenarios. The architecture model was primarily dedicated to set and handle merged environment for authority delegation mechanism, identity-based capability, and dynamic context information. This architecture supported the security management by applying the SOA technique depending on Web services. It promoted the usage of the SOAP and REST principles related to the characteristics and nature of devices and services. In such a situation, the SMGW manage all security features.

The robust security techniques have the following advantages: (i) A robust protocol can be one that does not break easily for the various attack models; that is, a security architecture in which any individual application can fail without disturbing the system or other applications can be said to be robust. (ii) Robust is also sometimes used to mean the architecture designed with a full complement of capabilities. Thus, in the context of IoT, the security protocols were designed for continuous operation with a very low failure rate.

2.3 Power-Efficient Techniques

One more major issue in security architecture development is power efficiency. In the literature, there are two significant works on power efficiency in security protocols. Urien et al. [49] proposed a protocol called LLCPS, i.e. the Logical Link Control protocol (LLCP) secured by TLS. LLCPS allowed an extensive choice of trusted services for the IoT in Near-Field Communication (NFC), which allows proximity communications with retiring throughputs (hundreds of kbit/s) and low power consumption (a few mW). The proposed LLCPS protocol was efficiently used for payment, transport, access control, or file transfer applications.

Altolini et al. [3] proposed the implementation and performance evaluation of security functionalities at the link layer of IEEE 802.15.4-compliant IoT devices. Particularly, the encryption and authentication mechanisms were completely implemented in software and used the hardware ciphers offered by the IoT platform. Furthermore, they explained all the characteristics of the implementation and, at last, they provided the experimental results to measure the performance of the consequent encryption and authentication scheme for certain implementation approaches. Furthermore, they provided the quantitative results on energy consumption, memory footprint, and the execution time of particular implementation approaches, and examined several significant trade-offs.

A power-efficient architecture has the critical advantage of less memory usage. This architecture has power-economic benefits, which have the advantages of heat and light, thus greatly saving money on electricity bills.

2.4 Time-Efficient Techniques

Computation time is a significant parameter for developing the security architectures for IoT. Some of the proficient security architectures for IoT security are reviewed here. Gu and Wu [13] introduced a mutual authentication protocol for RFID tags that complies with ISO 18000-6B standard. It was employed in the low-cost tags as it needs only around 1000 gates. By performing security and performance analysis, the protocol was accepted as an efficient one. Giuliano et al. [12] tackled the security issue for non-IP devices that are able to connect by a short range with a mediator gateway, which can be seen as a short-range extension of conventional access network, in order to efficiently capture the IoT traffic. Specifically, a security algorithm was presented for both uni- and bidirectional terminals, based on the capability of the terminals. The security algorithms depend on a local key renewal, and only the local clock time is considered to perform it. Performance was analyzed by considering the maximum number of terminals that can be managed by one mediator gateway and the maximum packet delay as a function of the number of terminals in the area.

Moosavi et al. [31] proposed the security architecture. In this architecture, distributed smart e-health gateways perform the authentication and authorization of a remote end user, so that the medical sensors are relieved from doing these tasks. The architecture is based on the certificate-based DTLS handshake protocol, as it was the foremost IP security solution for IoT. A prototype IoT-based health-care system is developed for testing the authentication and authorization architecture. The prototype was made up of a Pandaboard, a TI SmartRF06 board, and WiSMotes. The CC2538 module incorporated into the TI board works as a smart gateway, and WiSMotes works as medical sensor nodes. The architecture was very secured than a centralized delegation-based architecture, as a highly secure key management method is used between sensor nodes and the smart gateway.

Han et al. [14] proposed a simulator, called DPWSim, in order to aid the use of IoT. DPWSim featuring secure messaging, dynamic discovery, service description, service invocation, and publish-subscribe eventing can be utilized to prototype, develop, and test products in terms of DPWS communication protocols. It can also maintain the relationship between the designers, developers, and manufacturers in developing a product. Borgohain et al. [5] examined several authentication systems employed for improved security and private relocation of an individual’s login identifications. The first part described the multifactor authentication (MFA) systems, which may not be suitable to IoT but offers good security to a user’s identifications. Following MFA, a short explanation about the working principle of communication between the third-party clients and private resources on the OAuth protocol framework, and an examination about the delegation-based authentication system in IP-based IoT, is presented.

Farash et al. [9] presented the enhanced security architecture of user authentication and key agreement scheme (UAKAS). This design facilitated the similar function with enhanced security level and allowed the heterogeneous wireless sensor networks to grow dynamically without manipulating any party involved in UAKAS. The results of security analysis by BAN-logic and AVISPA tools proved the security properties of the scheme. de Fuentes et al. [7] dealt with the problem by providing the concept of Probabilistic Yoking Proofs and establishing three main measures to calculate their security, cost, and fairness. The message structure found in classical grouping proof constructions is combined with an iterative Poisson sampling process by the proposed solution. Here, the probability of each object being sampled changes with respect to time. A number of mechanisms were introduced by them to apply fluctuations to each object’s sampling probability, and they proposed several sampling strategies.

Savola and Abie [40] investigated the security objective decomposition methods for an IoT e-health application. These methods lead to the growth of significant security parameters that are relevant to the security, contextual, and threat changes, and they are important for patient-centric IoT solutions utilized in various environments. To utilize these benefits, a context-aware Markov game theoretic model was formulated for security metrics and the risk impact assessment, to noticeably assess and authenticate the run-time adaptivity of IoT security solutions. Mahalle et al. [29] developed the Identity Authentication and Capability based Access Control (IACAC) model with protocol valuation and performance evaluation. The concept of capability for access control was initialized to safeguard IoT from various attacks like man-in-the-middle, replay, and denial-of-service (DoS) attacks. This technique presented a hybrid approach of authentication and access control for IoT devices. The outcomes of other associated studies were been examined to verify the results. Finally, the protocol was assessed by using a security protocol verification tool, and the verification outcomes revealed that IACAC was secure against the above-mentioned attacks.

Hu et al. [18] proposed a mutual identity authentication and key update scheme that is employed in the IoT with multihop relay devices in the middle link. In the IOT paradigm, terminal nodes have inadequate computation capabilities, and the obtained information is sent to the layer using relay devices. Thus, by using the relay devices, the technique inflicts less computation and communication requirements on terminal nodes, increases the computational overhead, and also authenticates the relay devices in some way and recessively, thereby securing the terminal nodes and relay devices from being compromised. Lai et al. [24] presented a conditional privacy-preserving authentication with access linkability (CPAL) for roaming service, which provides universal secure roaming service and multilevel privacy preservation. CPAL presented a nameless user linking function by using a group signature technique, which does not only effectively conceal user identities but also facilitates the authorized entities to connect all the available data of the identical user without obtaining the user’s real identity.

The time-efficient architecture requires less computation time to perform the authentication process, which reduces the power and energy significantly.

2.5 Energy-Efficient Security Techniques

The security architectures of IoT related to energy-efficient techniques are reviewed in this section. He and Zeadally [15] explained about the required security measures of RFID authentication schemes, and provided ECC-based RFID authentication schemes in terms of performance and security. Even though many of them do not fulfill all the security needs and have acceptable performance, they established that the ECC-based authentication methods are relevant for the health-care environment in terms of their performance and security.

Vucinic et al. [51] presented OSCAR as an architecture for end-to-end security in IoT. It was dependent on the concept of object security that connects security with the application payload. The architecture integrated authorization servers that offer clients with access secrets that allow them to demand resources from constrained CoAP (Constrained Application Protocol) nodes. The nodes reply with the requested resources that were signed and encrypted. The scheme inherently supported multicast, asynchronous traffic, and caching.

Zhao [56] utilized the custom data packet encapsulation mechanism to reduce the cost of data resources. Depending on the cross-platform communication descriptions, joined with secure encryption and decryption, signature, and authentication algorithm, a secure communication system of things model for the differentiation of things communication environment, which provides a standard packet structure, called smart business security IOT application ISSAP (Protocol Intelligent Service Security Application Protocol), was established. Kozlov et al. [23] explained the overall architecture for IoT and discussed about the well-known and new threats to security, privacy, and trust at various architecture levels, with attacker-centric and system approaches.

Shafagh et al. [41] proposed a system for safely storing IoT data in the cloud database. At the same time, query processing on the encrypted data is allowed. In order to achieve this, they encrypted the IoT data with a set of cryptographic methods. To enable this on resource-limited devices, their system depends on optimized algorithms that speed up the partial homomorphic and order-preserving encryptions by 1–2 orders of magnitude. Peretti et al. [35] presented a scheme called BlinkToSCoAP, achieved by the combination of three software libraries implementing energy-efficient versions of the DTLS, CoAP, and 6LoWPAN protocols over TinyOS. Moreover, a complete experimentation was presented that examines the performance of DTLS security blocks. The experiments examined BlinkToSCoAP messages exchanged between two Zolertia Z1 devices, which allowed evaluations in terms of energy consumption, memory footprint, latency, and packet overhead.

A two-way authentication security scheme for the IoT was implemented by Kothmay et al. [22]. It depends on the previous Internet standards, particularly the DTLS protocol. This security scheme is dependent on the commonly used public key cryptography (RSA), and functions over standard low-power communication stacks. They thought that by depending on a well-known standard, the engineering techniques, existing implementations, and security infrastructure can be used again, which facilitated a simple security uptake. They provided an implemented system architecture for the security system depending on a low-power hardware platform that is appropriate for IoT.

The main benefits of an energy-efficient security architecture are that (i) it has less energy consumption and (ii) it is more performance oriented.

2.6 Reliable Security Techniques

Nowadays, reliable and flexible security mechanism becomes a major concern. Some of the latest approaches for IoT presented in the literature are discussed here. Lake et al. [25] explained about use case scenarios and a secure architecture framework. Gessner et al. [11] presented trust-enhancing security functional components for the resolution infrastructure as a fundamental part of IoT. A preliminary requirement analysis and a critical control points assessment lead to the trust-enhancing security functional components to cover the basic IoT resource access control (AuthZ and AuthN), and the necessary functions including key exchange and management, identity management, and trust and reputation management. This composition of components with their interdependencies offers secure communication among subjects to assure an unbreakable communication, and hence includes devoted features of data integrity and confidentiality, service trust, and privacy of users. Keoh et al. [20] presented a clear evaluation of solutions for secure communication in IoT, particularly the standard security protocols to be employed in combination with the CoAP, which is an application protocol particularly personalized for the requirements of adjusting the constrictions of IoT devices. As the DTLS has been selected as the channel security under CoAP, they also considered the new standardization efforts to improve the DTLS for IoT applications. This incorporates the use of (i) a raw public key in DTLS; (ii) enlarging the DTLS record layer to guard multicast communication; and (iii) profiling DTLS for minimizing the size and complication of implementations on embedded devices.

Xu et al. [53] introduced an architecture called TAEC (Trustworthy Agent Execution Chip), to use the high-security, cost-effective software and hardware platform for the safe operation of agents. This technique is for fixing TAEC on every sensor node, to provide autonomic trusted hardware execution environment for agents. Pohls et al. [36] formulated a framework to ensure a configurable balance between reliability and privacy by considering security and privacy mechanisms in their early design stage. The RERUM scheme comprises architecture built on new network protocols and interfaces, and the design of smart objects hardware. To emphasize the challenges and evaluating the scheme, RERUM used various Smart City applications that were installed and evaluated in the real-world test beds in the two Smart Cities that participated in the project.

A reliable architecture does not silently continue and deliver results that include uncorrected corrupted data. Instead, it detects and, if possible, corrects the corruption, for example by retrying an operation for transient (soft) or intermittent errors, or else, for uncorrectable errors, isolating the fault and reporting it to a higher-level recovery mechanism.

2.7 Smart-Level Security Techniques

Techniques that do not fall under the categories discussed above are reviewed in this section. Urien et al. [50] proposed a model for secure NFC services depending on the peer-to-peer (P2P) mode. NFC was a convenient communication technology, aided by smartphones or a consumer device that looks like a capable technology for the IoT. It was widely employed in several applications, for instance, transport, payment, access control, and most commonly for small information exchanges. LLCP manages the NFC P2P sessions. LLCPS was introduced as a TLS security layer working over LLCP. It imposes data privacy and integrity, and also provides identity to smart objects.

Maarten et al. [28] explained about a usage control toolkit for handling the above problems. The usage control toolkit describes the rules in managing the access to data and resources in IoT, i.e. the policies can be defined for different contexts such as work, personal life etc., and for different roles. Shah et al. [42] proposed a scheme to improve the available access control systems. This method of improving the access control system makes sure that the system is wireless, thus reducing the wiring problems. The explained prototype has the function of taking inputs from a smart card reader or a biometric sensor. These inputs were executed inside the controller (TM4C123GXL-based on ARM Cortex-M4).

Sivaraman et al. [47] focused on some of the smart-home appliances existing in the market, and considered their operation to expose various security and privacy issues. Implementing and practicing security in different devices vary based on several factors, such as device capabilities, mode of operation, and the manufacturer. This leads to the development of a design that employs additional security measures in the network. They used software-defined networking to implement dynamic security rules that can change depending on the situation, such as time-of-day or occupancy of the house. Park and Bang [34] employed the Jeju-VTS middleware to support information exchange on sea traffic. They framed a system that enables Inter-VTS Data Exchange Format (IVEF) service simulation in an IoT environment. To do this, they used an Android mobile phone and a personal computer for the emulation of a ship on cruise and VTS center. Alqassem and Svetinovic [2] presented security-relevant policies to reduce the identified forms of security attacks and to decrease the susceptibilities in the upcoming growth of the IoT systems. At last, it was employed on an IoT smart grid scenario.

Riahi et al. [38] investigated a technique for designing security mechanisms and its usage for IoT. They stated that the general method to security problems does not remove all the characteristics linked to this novel concept of communication, sharing, and actuation. Actually, the IoT concept includes new features, mechanisms, and risks that cannot be entirely considered through the classical formulation of security problems. Torjusen et al. [48] incorporated runtime verification enablers in the feedback adaptation loop of the ASSET adaptive security framework in order to achieve self-adaptive security and privacy properties in the e-health settings. The enablers make machine-processable formal models of the system state and context presented at run time, make requirements to describe the purposes of authentication, and enable dynamic environment supervision and adaptation.

The benefits of the IoT enable humans to control, effectively and easily, things that are up close or remotely. For example, the user can manage his car engine and control it from his computer. However, the best case is when different “things” communicate with each other using the Internet Protocol. For example, a refrigerator can communicate with a shopping center and buy supplies without human intervention.