244

S. Das et al.

From a financial point of view, efficiency is the most universally important aspect of

any technology for any industry. The efficiency of a blockchain in any system can

differ significantly depending on the type of consensus algorithm and cryptographic

hashing algorithm used as well as the computational power of the IIoT device in

question. All cryptographic hashing functions have different complexity, levels of

provided security and often have their own vulnerabilities.

The aim of the proposed model is to create a system that caters to a range of

devices with varying computational powers, instead of going with the common route

of trying to find “a size that fits all”. So, we have chosen a context-aware solution that

dynamically adjusts to the specific needs of the device and the computation power

available. To achieve this, a benchmarking algorithm is designed, which is used to

estimate the computational power of various IoT devices in the network. This bench-

marking algorithm consists of multiple trigonometric functions that are repeated over

theentire360°rangeanditeratedmultipletimesinafixedamountoftime.Whilethere

are other functions like matrix manipulation functions and prime number generation

functions that are used by some other benchmarking algorithms, these can overwhelm

the weaker processors of some low-power IIoT devices resulting in inconsistencies.

Trigonometric functions are very complex and the sheer number of calculations can

stress the processors of a wider range of devices without any inconsistencies. As the

time of operation is fixed, the number of iterations achieved within the fixed time

gives a fair indication of the computational power of the processor used. Compared

to other methods of processor benchmarking that can take hours on weaker hardware,

this is a much more time-efficient process as the fixed time can be set to very low.

The average number of iterations achieved is the output benchmark score.

The output score from the benchmarking function is used to generate a device-

specific token, which is stored in the device for future reference. This token contains

the benchmark score achieved by the device and is used to categorize the huge range

of possible IIoT devices into multiple tiers. In our case, the devices are classified into

four tiers, Tier I to IV, with ascending computational power, where Tier I consists of

low-power devices and Tier IV consists of computationally high-power devices.

Now, the IoT network is split into numerous virtual clusters of IIoT devices,

according to their tier. Every tier is accordingly assigned a unique hashing algorithm,

which is most suited for the processing power of its devices and is most appropriate

for the data it is storing. When the consensus algorithm is called, the token of the

concerned device is checked to find its tier as shown in Fig. 1. Its respective hash

function is used to generate the hash. As such, we have various parallel blockchains

in place all with their unique hashing algorithm in the entire IoT network. This

makes the overall system more secure as it is difficult for attackers to identify the

hash algorithm being used, with the IoT network separated into different clusters

of IIoT devices as demonstrated in Fig. 2. By having a system in place that can

assign different hash functions, we also solve the problem of not having enough

crypto agility partially, as this architecture allows any vendor to add or update their

blockchain to any hash function as needed at least from the software side as long as