1. Introduction
About 500 billion devices will be connected to the Internet by 2030 and this represents a tenfold increase based on the 50 billion forecast for 2020 [
1]. The traditional approach for implementing the Internet of Things (IoT) systems is the centralized architecture. In such an architecture, the IoT devices send information to the cloud server or a central entity, which then processes the information. The deployment of billions of IoT devices brings some fundamental design challenges with the centralized architecture in terms of privacy, security, confidentiality, scalability, slow operation, and high cost when third-party authentications are required [
2,
3].
As a fundamental technology for cryptocurrencies, blockchain technology can be used to address the security and trust vulnerabilities, and the high maintenance cost associated with the conventional IoT networks [
4,
5,
6]. In particular, blockchain technology can be used in IoT device tracking, coordination and for processing transactions (The terminology “transactions” refers to any value information exchange between the nodes in the network) between devices. The decentralized blockchain network mitigates the risk of a single point of failure in centralized systems with nodes that transfer data in a peer-to-peer manner [
6,
7,
8]. Moreover, an autonomous IoT network can be actualized when IoT devices or ‘nodes’ communicate in a distributed manner [
9]. This requires a consensus method that allows nodes to agree on the validity of the message or communicated data. On the other hand, the blockchain system utilizes a consensus mechanism to agree on new data [
10,
11,
12]. Hence, consensus mechanisms that are used in blockchain can also be applied to IoT networks. However, most IoT devices are constrained in terms of energy since they are battery-powered. They are also in limited bandwidth resources and computational capabilities. The aforementioned factors must be taken into consideration when selecting consensus mechanisms for IoT networks [
8,
11,
13].
1.1. Practical Byzantine Fault Tolerance: A Consensus Mechanism for IoT Networks
The growth of Internet of Things (IoT) networks presents substantial challenges in maintaining data integrity, security, and fault tolerance across distributed, often resource-constrained, devices. Traditional centralized architectures struggle with scalability and single points of failure, making decentralized approaches, such as blockchain, appealing for IoT applications. Practical Byzantine Fault Tolerance (PBFT)—a blockchain consensus mechanism—offers low computational power and complexity, and it is well suited for IoT networks [
13]. PBFT is a practical and improved protocol on Byzantine Fault Tolerance (BFT) that was proposed and implemented in [
14], and it has better response time and peak throughput than BFT. It achieves these gains by incorporating several important optimizations, including reducing the size and number of messages, using message authentication codes, and integrating incremental checkpoint-management techniques. PBFT provides a significant reduction in energy consumption when compared with other blockchain consensus mechanisms such as PoW (proof of work) [
15]. It is not affected by the “nothing at stake” and the centralization problems associated with PoS (proof of stake) [
13]. The PBFT protocol guarantees safety and liveness when no more than
out of the total
n nodes are faulty. This implies that neither software or operator errors nor alterations by adversaries can cause a crash in the system when the number of faulty nodes is lower than this threshold.
PBFT has three main parts namely, the normal PBFT, the garbage collection, and the view change. The purpose of the normal operation is to ensure that the client’s (IoT device) requests are executed in a predefined order. It is the main procedure for making agreements among the nodes. The purpose of garbage collection is to recover in case of an accident. The view change purpose is to select a new primary (PBFT consensus mechanism selects one of the nodes in the network as the primary node while other nodes serve as backups) node and it takes effect when the primary in the consensus network either becomes a faulty node or breaks down. In the normal operation of the PBFT, when the primary node of the PBFT network receives the client request, it starts the three phases of the PBFT consensus mechanism namely, pre-prepare, prepare and commit. In the pre-prepare, the primary broadcasts a pre-prepare message to other nodes within the PBFT network. Moreover, in the prepare and commit phases, nodes communicate with each other using the broadcast protocol. By using these inter-node communications, the PBFT consensus mechanism ensures that all nodes are synchronized and agree that the transactions are legitimate, which are then added to the blockchain. PBFT consensus is thus critically important to blockchain systems since the consensus mechanism largely determines the blockchain system performance in terms of transaction throughput, delay, scalability, security, etc.
1.2. Motivation and Related Work
The majority of IoT devices rely on wireless communication protocols like WiFi and cellular networks, with blockchain integrated to enhance security and trust. Nevertheless, wireless communications between nodes face unique challenges due to inconsistent channel quality, network congestion, and inherent openness, which can impact the security, reliability, and performance of blockchain systems. Traditional PBFT (Practical Byzantine Fault Tolerance) networks generally assume ideal communication conditions, free from latency and bandwidth limitations. However, in wireless environments, fluctuating channel strength, restricted channel resources, and varying network topologies introduce complexities that must be addressed to understand how wireless communications affect PBFT-based blockchain network performance.
Previous research has explored alternative consensus mechanisms in wireless blockchain environments. For instance, Cao et al. [
12] studied a Tangle consensus model in a wireless blockchain network. Further, Xu et al. [
16] analyzed the security performance of Raft-based wireless blockchain systems. The authors in [
4] developed a framework for evaluating blockchain transaction throughput and node placement in Proof-of-Work (PoW) based wireless networks. However, the analysis presented in [
4,
12,
16] does not apply to wireless PBFT consensus mechanism due to its 3-phased operation and its peculiar inter-node communications.
Recently other BFT protocols such as scaled BFT (SBFT) [
17], FastBFT [
18] and HotStuff [
19] have emerged. These new protocols have reduced communication cost while maintaining the safety, liveness and resilience properties of PBFT, thus making them more suitable for IoT networks. Though the analysis presented in this paper focuses on the wireless PBFT protocol, it can be easily extended to other wireless BFT protocols such as SBFT, FastBFT and HotStuff.
1.3. Contributions and Organization
In this paper, we propose a novel framework for implementing the PBFT consensus mechanism over a wireless channel. In particular, we consider that communications between the IoT nodes in the three phases of the normal operation of PBFT (i.e., pre-prepare, prepare and commit phases) are done over the wireless channel where node contend for channel access based on the slotted ALOHA protocol [
20]. We derive the success probabilities over the prepare and commit phases of the wireless PBFT network for the case with non-faulty primary (The pre-prepare phase success probability is equivalent to one since the primary node is non-faulty and there are no collisions in the pre-prepare phase). Besides, we derive the end-to-end success probability while considering an ideal pre-prepare and finishing phases (The nodes send their reply to the client in the finishing phase). The end-to-end success probability serves as the basis for deriving other wireless PBFT key performance indicators (KPIs) namely, the transaction throughput which is expressed in terms of the number of transactions per second, transaction confirmation delay, the optimal transmission interval, and the viable area. We show the impact of view change as a result of the faulty primary by deriving the effective transaction confirmation delay. The transaction throughput demonstrates the transaction processing capacity of blockchain system and it’s maximized at the optimal transmission interval. This optimal interval also achieves the minimum end-to-end delay and it can be used to set the PBFT network timeouts. On the other hand, the viable area defines the minimum PBFT network’s coverage area required for a successful implementation of the PBFT consensus mechanism. In other words, the viable area ensures that the minimum number of nodes is activated in the wireless PBFT network, while satisfying all constraints. This will lead to significant savings in energy and an improvement in the overall performance. The viable area is decided by parameters such as the nodes’ transmit power and receiver sensitivity, transmission interval
v, propagation channel parameters, success probability requirement and the number of faulty nodes
f.
The main contributions of this work are summarized as follows.
Considering the Spatio-temporal domain Poisson point process (PPP) modeling, i.e., node geographical distribution and transaction arrival rate in the time domain, we establish an analytical framework for obtaining the end-to-end success probability of the wireless PBFT network. Our framework provides a mathematical link between the wireless PBFT network’s coverage, the transmission interval, the nodes transmit power and the number of faulty nodes in the network. The proposed framework enables the optimization of network parameters such as the transmission interval, the end-to-end transaction throughput and delay, the broadcast transmit power and the coverage range.
Using the framework, we define the viable area of the wireless PBFT network, which ensures that the minimum number of nodes are activated for a given end-to-end success probability requirement while guaranteeing the networks liveness, safety, and resilience. Furthermore, we present a low complexity algorithm for obtaining a viable area of the wireless PBFT network.
We derive the effective transaction confirmation delay to show the effect of view change on the wireless PBFT network. Furthermore, numerical results are presented to verify our analytical derivations.
The rest of the paper is organized as follows. In
Section 2, we present the system model and the description of the wireless PBFT algorithm.
Section 3 presents the wireless PBFT network including its implementation constraints and the analysis of the independent prepare and commit success probabilities, and the end-to-end success probability.
Section 4 presents the performance metrics of the wireless PBFT network namely, the average transaction throughput and delay, optimal transmission interval, effective transaction confirmation delay, average transmit power and the viable area. Numerical and simulation results are presented in
Section 5. Finally,
Section 6 concludes the paper. A preliminary version of this paper has been reported in [
21] where we introduced the viable area concept for the wireless PBFT network without considering the effect of channel contention. Herein, our analysis has been made more generic and realistic with the inclusion of the effect of channel contention. Furthermore, new performance metrics such as the end-to-end success probability and the average transaction throughput have been derived. A detailed analysis of the optimal transmission interval has also been conducted.
3. Wireless PBFT Network
IoT is one of the main application scenarios of blockchain [
24], and IoT nodes could be connected wirelessly. Considering the unstable channel quality, interference, limited resource, and various network topology [
25], it is necessary to investigate the impact of wireless communications on the PBFT-based blockchain networks. We consider that the nodes are spatially distributed in
according to a homogeneous PPP with density
. Furthermore, we consider a noise-limited wireless network where all nodes have the same receiver sensitivity
. In the first phase of the wireless PBFT consensus system, the primary node broadcasts a pre-prepare message to the whole network over the wireless channel, as illustrated in
Figure 2. We consider that all nodes are equipped with an omnidirectional antenna. Hence, the coverage range of the primary node can be expressed from [
26] as
where
is the transmit power of the primary node. Note that the coverage range is based on the maximum long-term averaged channel power, i.e., the effect of the channel fading has been averaged out. The parameter
is the pathloss exponent,
is a reference distance for the antenna far-field, and
V is a unit-less constant that depends on the antenna characteristics and the average channel attenuation. The broadcasted pre-prepare message by the primary node will be received by all other nodes within its coverage radius, i.e.,
. These nodes then perform the necessary verification to ascertain the validity of the block, as shown in
Figure 1. The number of nodes within the coverage of the primary can be obtained from the following approximation
where
is the expected number of other nodes within a circle of radius
centered on a typical point of the process, i.e. the primary node. Further, the
K function
for
.
Constraint 1
In the wireless PBFT consensus network with
f faulty nodes, the number of nodes within the primary node’s coverage, which is denoted by
n, must satisfy the expression in (
1). The primary node’s coverage, which is also within the network’s coverage, can be defined as
. Hence, the constraint can be expressed as follows:
Each node then enters the prepare stage after accepting the pre-prepare, as illustrated in
Figure 1. Note that a larger primary node coverage will increase the safety of the network but this also leads to an increase in the energy consumption and besides, an increase in the likelihood of collision of messages broadcasted by the nodes in the prepare and commit phases.
3.1. Channel Contention in Wireless PBFT Consensus Network
We utilize the slotted ALOHA as the medium access protocol for the wireless PBFT-based blockchain network. Moreover, LoRa is a low-power wide-area network (LPWAN) technology that has been specifically developed for IoT and it uses slotted ALOHA as its multiple access technique [
27]. In this model, we consider the PBFT nodes to be class B, and the primary may act as the central server. The length of the receiving window for the nodes is supposed to be slightly more than the combination of the timers of the prepare and commit phases. When the primary node receives a request from the client, it broadcasts a beacon to the other nodes to trigger the receive windows.In slotted ALOHA, time is divided into slots of size
and nodes start to transmit frames/messages only at the beginnings of slots. A collision occurs if two or more nodes transmit their message or packet at the same slot. Otherwise, there are no collisions and packets are properly sent as illustrated in
Figure 3a. Message broadcast in wireless PBFT network is done over the wireless channel where nodes contend with each other for channel access. We assume that packets are lost when there are collisions over the channel such that the sent messages are not received by other nodes in the network. Consequently, collisions have a negative impact on the liveness, safety, and resilience of the wireless PBFT consensus network. In particular, all nodes need to contend the same channel, thus collisions can occur which could then lead to a less secured consensus network.
The prepare and commit success probabilities and end-to-end success probability of the conventional PBFT network relies on the number of Byzantine/faulty nodes and the total number of nodes in the network. In addition to these two elements, the wireless PBFT consensus network must incorporate the effect of packet collision as well when evaluating its success probabilities.
3.2. Wireless PBFT Success Probability
In the following, we analyze the success probabilities of each phase of the wireless PBFT network and its end-to-end success probability.
3.2.1. Pre-Prepare Phase
There are no channel contention or collisions in the pre-prepare phase since only the primary node accesses the channel. Hence, the broadcasted pre-prepare message by the primary is successfully received and decoded by all nodes that are within the primary’s coverage as defined by (
3). Here we assume that the primary node of the current view is non-faulty. A faulty primary will warrant a change of view, i.e., selection of a new primary node.
3.2.2. Prepare Phase
Given that
nodes accept the pre-prepare message from the primary, then each node enters the prepare phase by broadcasting the prepare message to all other nodes. Hence,
nodes contend for channel access during the prepare phase. Note that the primary (denoted by 0) does not broadcast any message during the prepare phase, as illustrated in
Figure 1 and
Figure 3b. Here we assume that each prepare message is contained in a packet of length
, which is equivalent to the slot size, and that the population of nodes attempts to broadcast according to a Poisson distribution. Let the interval
denote the time frame between the first packet and the last packet in the prepare phase, as illustrated in
Figure 3b. Consequently, the traffic generated in the interval
can be expressed as
Considering the contention of nodes without the PBFT constraints in the prepare phase, the throughput in the prepare phase can be obtained as the conventional slotted ALOHA throughput
, i.e., the product of the rate of transmission
and the probability of success
[
20]. In that case, the optimal traffic load,
, such that the optimal interval
can be expressed as
. For non-faulty nodes to attain the prepared state in the wireless PBFT network, at least
transmissions by the nodes in the prepare state must be successful [
14]. Upon defining the probability of generating
i successful transmission over
contending node in a generic slot:
the success probability in the prepare phase can be obtained from the following Lemma.
Lemma 1. The success probability of the prepare phase of wireless PBFT network with slotted ALOHA access protocol can be expressed aswhere is defined in (5). Proof. To become prepared, a non-faulty node should have
prepare messages from different nodes that match the pre-prepare message from the primary node. Here we consider that the primary of the current view is non-faulty and that only the primary node transmits in the pre-prepare phase. Consequently, the transmission by the primary node in the pre-prepare phase will be successfully received by all nodes within its coverage. Now consider that each node broadcast its prepare message once and that there is no acknowledgement (ACK), request to send (RTS) or clear to send (CTS) messages due to the broadcast nature. Consequently, the number of transmissions in the prepare state is equivalent to
, i.e., the number of nodes excluding the primary. Note that only prepare message received from faulty nodes will not match with the pre-prepare from the primary. Further, since a log of
prepare messages from different nodes are required, the minimum number of nodes with successful transmission in the prepare phase is also set to
. Hence, the starting point of the summation in (
7) is
, and
. Note that
f out of the
(required minimum number of successful transmissions in prepare phase) successful transmissions could be transmissions from Byzantine or faulty nodes. Nevertheless, the prepared state will still be attained by each of the non-faulty nodes in this case since the received transmissions from the other
f nodes will match with the pre-prepare and achieve a majority of
as against the
f Byzantine broadcasts in the prepare consensus. The resilience of the PBFT consensus protocol is thus preserved by restraining the minimum number of successful transmissions in the prepare phase of the wireless PBFT to
, irrespective of whether Byzantine nodes’ transmissions have suffered from collision or not. □
Assumption (Non-Overlapping Prepare and Commit Phases)
We assume that all non-faulty nodes are prepared at approximately the same time instance such that traffic generated during the prepare phase does not overlap with the commit phase.
In Lemma 1, all nodes that are within the primary’s coverage broadcast a prepare message within the time-frame denoted
, and successful transmission from each node are received and decoded by all other nodes within the primary’s coverage. Consequently, all non-faulty nodes will be prepared, i.e., receive
prepare messages that match with the pre-prepare, at approximately the same time instance including non-faulty nodes that are yet to broadcast their own prepare message. Prepared non-faulty nodes thus move directly to the commit phase at the same time instance, and hence,
the prepare and commit phases of wireless PBFT consensus mechanism are non-overlapping. For tractability sake and to retain the resilience associated with the conventional PBFT protocol, the commit phase commences when
elapses, as shown in
Figure 3b.
3.2.3. Commit Phase
The operation of the commit phase in the wireless PBFT network is very similar to the prepare phase. Considering that
nodes were successful in the prepare stage where
, then
nodes, i.e., all successful nodes in the prepare phase and the primary node will contend for channel access in the commit phase. Each node broadcasts a commit message to other nodes when its prepared becomes true. We denote the interval between the first packet and the last packet in the commit phase by
such that the traffic generated can be expressed as
Similar to the prepare phase, the success probability in the commit phase can be obtained from the following Lemma.
Lemma 2. The success probability of the commit phase of wireless PBFT network with slotted ALOHA access protocol can be expressed aswhere is defined in (8). Proof. The minimum number of successful transmissions for a successful commit is
, which is the starting point of the summation in (
9) such that the commit success probability
. The rest of the proof follows directly from Lemma 1. □
3.2.4. End-to-End Success Probability
Given that the primary is a non-faulty node, the success probability of the pre-prepare phase is thus equivalent to one since there are no channel contention or collisions in the pre-prepare phase. Moreover, the success of the commit phase is dependent on the success of the prepare phase. Considering that the finishing state i.e., “reply to the client” by committed nodes is ideal (i.e., all successful commits are successfully reported to the client), the end-to-end success probability of the wireless PBFT network is given in the following Theorem.
Theorem 1. The end-to-end success probability of the wireless PBFT network with slotted ALOHA access protocol can be expressed as Proof. The proof follows from the dependence of the commit phase on the prepare phase, and Lemmas 1 and 2. □
From (
10), it can be seen that for
, the end-to-end success probability
when
and
.
Corollary 1. The end-to-end success probability of the wireless PBFT with the same transmission interval for the prepare and commit phases can be computed as Note that the end-to-end success probability is the probability that the transaction initiated by the client IoT device is agreed on by the nodes in the wireless PBFT consensus mechanism and added to the blockchain. The end-to-end success probability is thus a very important metric in the wireless PBFT network as it directly relates to KPIs such, as the transaction throughput and the end-to-end delay. In the next section, we utilize the success probability of the wireless PBFT network as the basis for deriving other KPIs.
5. Numerical Results
In this section, we present some numerical results to illustrate our analytical findings. The system parameters are as follows:
,
and
. Regarding the broadcast message, the channel bit rate is
, the payload length is
while the MAC and PHY header lengths are
and
, respectively, such that
[
29]. We set the PBFT coverage radius
and show results for the wireless PBFT consensus networks.
In this section, we evaluate the performance of the wireless PBFT network in terms of the optimal transmission interval , the end-to-end success probability and the end-to-end transaction throughput . We consider that the primary node is located at the origin and its coverage range (and equivalently the primary node’s transmit power ) is obtained by assuming that all nodes are with the same receiver sensitivity . and equivalently is defined such that there are nodes within the coverage of the primary. Note that required to achieve n can be adjusted based on the node density , however, the latter is fixed in our evaluations unless otherwise stated. From (3) and (5) the primary node’s transmit power and the number of nodes in the wireless PBFT network n is such that .
In
Figure 4, we compare the derived end-to-end success probability with simulation for the wireless PBFT network with
nodes and
faulty nodes. Results show a tight match between the analytical derivation and simulation results. As expected, it can be seen that for a fixed transmission interval, increasing the number of faulty nodes
f leads to a reduction in the end-to-end success probability since the likelihood of reaching the prepared or committed state reduces as
f increases. On the other hand, for a fixed number of faulty nodes
f, increasing the transmission interval increases the end-to-end success probability. This is due to the fact that the probability that transmitted prepare/commit messages/packets encounter a collision reduces with increasing transmission interval
v.
In
Figure 5, we show the effect of increasing the number of nodes in a fixed area (i.e., increasing the node density
) for
and transmission interval
. It can be seen that when the transmission interval is high enough such as
, increasing the number of nodes lead to an increase in the end-to-end success probability. This is due to the fact that the probability of a node receiving prepare or commit messages from non-faulty nodes increases as the node density increases. Moreover, when a particular number of nodes is reached, the effect of the collision of the nodes’ transmissions becomes more pronounced, and the end-to-end success probability starts to decrease as seen for the case with the transmission interval
.
In
Figure 6a,b, we plot the average end-to-end transaction throughput against the transmission interval for
(
) and
(
), respectively. Results show that there exists an optimal transmission interval that maximizes the average end-to-end transaction throughput.
Figure 6a further shows that increasing
f increases the optimal transmission interval
while reducing the optimal end-to-end transaction throughput. In addition,
Figure 6b shows that a larger PBFT network such as the case with
requires a much larger transmission interval than the case with
in order to achieve the optimal end-to-end transaction throughput
. The parameters
and
can serve as input when setting the PBFT network’s transaction timer. Note that at the expiration of the timer without a reply from the nodes, the client resends the service request to the primary node of the PBFT network.
In
Figure 7a,b, we show the effect of increasing the number of nodes
n on the optimal transmission interval and the end-to-end transaction throughput, respectively, for
. The sawtooth shape of the curves of
Figure 7a,b are due to the fact that
. At instances
, the corresponding
f increases, i.e.,
and the optimal transmission interval increases as seen earlier in
Figure 6. On the other hand, at instances
which all corresponds to
, the optimal transmission interval reduces. This is due to the fact that the probability of a node receiving prepare or commit messages from non-faulty nodes increases and out-weighs the effect of the collision as a result of the additional non-faulty node.
In
Figure 7a, we compare the optimal end-to-end transmission interval
with the optimal intervals obtained for the independent commit and prepare phases, i.e.
and
, respectively. It can be observed that the optimal end-to-end transmission interval
exceeds the optimal values for independent prepare and commit phases. Further, the optimal transmission interval for the commit phase exceeds that for the prepare phase, i.e.
. This observation is due to the fact the number of contending nodes for channel access in the commit phase exceeds that of the prepare phase since the primary does not broadcast in the prepare phase but the pre-prepare and the commit phases. Results in
Figure 7b shows that the end-to-end based approach results in much higher transaction throughput, i.e., more transactions can be added to the blockchain. Moreover, we also benchmark our result with the case with fixed transmission interval, i.e.,
(which is the optimal transmission interval for the case with
). As can be seen in
Figure 7b, using a fixed transmission interval is sub-optimal as it leads to much lower transaction throughput, and hence, demonstrates the need for the optimization of the transmission interval.
In
Figure 8, we show the results for the transaction confirmation delay when the effect of view change has been integrated. We plot the transaction confirmation delay without view change
, the average view change delay and the effective transaction confirmation delay
for the case with
. The transaction confirmation delay without view change was obtained from (16). The results shows a tight match between the simulated and theoretical results. It can be seen that the view change leads to a significant increase in the effective transaction confirmation delay
. Furthermore, increasing the number of faulty nodes
f also leads to an increase in the effective transaction confirmation delay.
In
Figure 9, we plot the viable area of the wireless PBFT network against the transmission interval
v for a fixed end-to-end success probability
and number of faulty nodes
f. For the case with
and
, it can be seen that the wireless PBFT network becomes unviable for a transmission interval
where
. Moreover, this constraint can be relaxed by either setting a more relaxed end-to-end success probability requirement as in the case with
or reducing the number of faulty nodes as in the case with
. In other words, a much lower transmission interval
v is required for the wireless PBFT network to become viable when either the end-to-end success probability or the number of faulty nodes is lowered. Further, for the case with
and
, it can be observed that once the network becomes viable, the viable area and equivalently the number of nodes
n reduces as the transmission interval
v increases. Moreover, the viable area converges to the effective area of the minimum number of nodes required to achieve
constraint 1, i.e.,
with a further increase in the transmission interval
v.
6. Conclusions
In this paper, we investigated the performance of the PBFT protocol when implemented over the wireless network. We first introduced the wireless PBFT network system model and the channel contention assumptions. The success probabilities for the prepare and commit phases, and the end-to-end success probability have been derived for the wireless PBFT network. Using the end-to-end success probability as the basis, we derived the expression for other KPIs of the wireless PBFT network namely, the transaction throughput, transaction confirmation delay, the optimal transmission interval, the average node transmit power and the viable area. We also derived the effective transaction confirmation delay to show the effect of the view change on the wireless PBFT networks. Numerical results validated the accuracy of our theoretical analysis. Furthermore, the viable area of the wireless PBFT network achieves liveness, resilience, and safety with the minimum number of nodes, minimum transmission interval, and minimum broadcast transmit power and thus results in significant energy savings.
Note that the analysis of the wireless PBFT network presented in this paper was based on the static IoT network scenario. Since most IoT devices intrinsically work over mobile systems or evolve toward mobility, performance analysis of the wireless PBFT networks that capture mobility deserves attention in future studies. In addition, the analysis only considers the collisions that might affect the system’s effectiveness and efficiency of the consensus. Therefore, the impacts of malicious attacks on the security of the system are also worth investigating in future studies. Moreover, the implementation of other wireless protocols for the PBFT network is also worth exploring.