Computational Cost. In this subsection, the proposed scheme is compared with the schemes proposed by Xxx et al. [51], Xxxxxx et al. [53], Xxxxx et al. [42] and Xxxxx et al. [54]. The key results obtained from the comparison are shown in Table II to determine the effectiveness of the proposed scheme. The TABLE III COMPARISON OF COMPUTATIONAL COSTS (IN MILLISECONDS) Schemes 𝒟ℛ𝒩𝑢/smart IoT device 𝒟ℛ𝒩𝑣/smart IoT device Total (in milliseconds) Das et al [51] 5.28 5.28 10.56 Xxxxxx et al [53] 5.28 5.28 10.56 Xxxxx et al [42] 8.62 18.65 27.27 Xxxxx et al [54] 5.28 5.28 10.56 Proposed 2.88 2.88 5.76
Computational Cost. As can be seen from the Table 1 and Figure 4, there is a huge advantage in computation costs. With respect to computation costs, the hash operation is the most efficient operation, compared to elliptic curve multiplication, elliptic curve addition or encryption. Table 1 Comparison of the computation costs. (Tm=Time for EC point multiplication, Ta=Time for EC point addition, Ts=Time for encryption, Th= Time for hash operation Entity [36] This Device MEC TTP 7Tm + 2Ta + 2Ts + 00Xx 0Xx 0Xx + 2Ta + 2Ts + 00Xx 0Xx 0Xx + 4Ta + 4Ts + 13Th 3Th By using the average computational time in [35] for different cryptographic operations in different devices, we compare the average computation time in a practical implementation. Here we assume Zolertia RE-xxxx as the MIoT node, Raspberry PI3B+ as the MEC node and PC Intel Core i7-8750HCPU as the TTP or the cloud. As can be seen from Figure 4, the proposed solution has a significant performance advantage over the system proposed in [36].
Computational Cost. No expensive and major operations like ECPM and M-Exp are involved in our proposed scheme. In designed scheme [12], four ECPM and two M-Exp operations and in scheme [8] two M-Exp are used. Graph in Fig. 4 shows that our scheme is efficient in computation cost as compare to [12], as compare to [8] and as compare to [23]. In our scheme we implement the experiment done in [24] on MICA2 sensor that is operational with low power ATmega128 8-bit micro-controller at 7.3728 MHz, 128 KB nonvolatile memory ( ROM) and 4 KB volatile memory ( RAM). One major operation ECPM uses 0.81s using 160 bits elliptic curve [25] and RSA 1024 bits M-Exp takes 22 seconds [26]. DES encryption and decryption execution time [27] is same which 4.543859 seconds. We calculate the computation cost of our scheme in comparison with the [8], [12] on the basis of the results of [23], [24], [26]-[28]. According to scheme [28] the 3rd generation MICA2 needs 2.66s for pairing computation. The computational time of our proposed scheme is negligible as compared with others existing schemes [8], [12] because we used symmetric algorithm for encryption and decryption as well as our scheme is more suitable for resource constraint environment of BSN. One ECPM operation consumes 19.1Mj and one pairing computation operation consumes 62.73mJ energy [24], [28]. Our scheme have no major operation so energy consumption as compared to others existing schemes is negligible. TABLE II. ASSOCIATED PARAMETER AND DATA SIZE INVOLVED IN DATA COMMUNICATION Associated Parameters Data Size RSA Key 1024 bits ECC Key 160 bits DES Session Key 64 bits Master Secret Key 128 bits Sensor ID 48 bits 16 bits Comparison of Computation Cost of existing and proposed
Computational Cost. When comparing the results obtained between experiments and various numerical models, one has to keep in mind the cost associated to each of these numerical approaches. Indeed, this parameter varies quite significantly depending on the considered approach and it may orient the decision to prefer one to another. Therefore, this section aims at comparing it for the three methods of interest in this paper. It will enable to add context to the presented results and analyse the obtained results through a new angle. It includes the time necessary to go from the definition of the model to the post-processed data. The time taken for each simulation technique is mentioned step-wise in Table 11 and Table 12. Table 11 indicates the time taken for the simulations with only current flow. Table 12 indicates the time taken for the simulations with current and waves. For time domain numerical simulations, the indicated time is the time taken to attain a converged solution for 30 seconds of time-series data. The values are provided based on the Edinburgh generic turbine running at TSR = 7 in order to ensure comparable information. For the flow only condition, the time presented in Table 11 is for 12 computational seconds of the BEMT-CFD model. The time provided is only indicative and meant to provide the reader a guidance about the scale of compu- tational effort required for each simulation technique. It may also be noted that for this comparison study, all numerical models are assumed to be ready-to-exploit and bug-free, i.e. the time associated with the develop- ment of the numerical code and bug fixing are not included. Also, the time expend for re-running of simulations is also not included. Finally, the total simulation time has to be put in perspective of the CPU specifications used for each code. It represents the parallelisation capacity applied on the calculation process to reduce its computation time. As such, even if BEMT-CFD and Blade-Resolved CFD have similar total simulation time, Blade-Resolved CFD numerical model has a significantly higher cost. Table 11: Comparison of computation time for different models (Edinburgh generic turbine at TSR = 7 - Flow only case) Event BEMT BEMT-CFD Blade-Resolved CFD Specification Duration - (Convection of the wake along the domain) ° Pre-processing 20 mins 9 hours 24 hours Simulation Computation 5 secs 2 days for 12 s 3 days Time Post-processing 5 mins 2 hours 2 hours Total Time 25 mins 2.5 days 4 days CPU Processor Intel ...
