Methodology

As mentioned in the previous sections, Reliability, and therefore the CDF function can be used to calculate the probability of survival or failure for a given system respectively. The CDF, can be used to describe the probability that the system will finish its operation prior to a given threshold and therefore indicate the impact of its subsystems’ variation on the total system lifetime. In security analysis, the CDF function can easily be adopted and treated as a quality factor that describes all encryption parameters and their impact on energy consumption. It serves as an indicator of the performance of the encryption parameters with respect to the energy consumption of the overall security system. This forms the basis for the proposed adaptive security scheme that extends the fitting of the model for each security mode accordingly, by properly adjusting functional parameters and always taking into consideration the energy cost. In this way, a metric that indicates the impact of all encryption parameters is developed, and thus a global indicator is derived. The proposed model can be thus considered global, as it is not based on distinct parameters, but, instead, arises from the impact of all the individual encryption parameters on energy consumption [4]. According to Equation (2.5), to calculate system reliability one should isolate all encryption parameters and calculate their individual probabilities P(T1 > t), P(T2 > t), P(T3 > t), P(T4 > t) accordingly. However,

it is not easy to derive their individual encryption parameter distributions since these cannot be isolated. Therefore, in this work, several simulation runs have been performed, as described in section 3.2.1, and the results provided the means to determine the empirical CDF (ECDF) of 𝑃(𝑇 > 𝑡). The encryption times of the 576 cases considered in this work have been measured and the ECDF for each security mode has been concluded. Based on the ECDF, the security modes can be compared

56

and evaluated. Depending on the requirements, the selection of the most efficient mode can be made either based on the security restrictions or the energy/time threshold.

Variables: list Cases # 576 cases data set

boolean S # security requirements

boolean E # energy requirements

list Sr # security restrictions

float Er # energy restriction

float R # reliability

float Rmin = 1 # lowest reliability

int efficient = 0 # most efficient case

1: if S = TRUE then

2: for case in Cases

3: if case not in Sr then

4: delete case 5: else 6: calculate R 7: if R < Rmin 8: R = Rmin 9: efficient = case 10: end if 11: end if 12: end for 13: return efficient 14: else

15: for case in Cases

16: if case not in Er then

17: delete case 18: else 19: calculate R 20: if R ≠ 0 21: delete case 22: end if 23: end if 24: end for 25: return Cases 26: end if

In the first case, the security modes that do not meet the security requirements are Figure 4.2: Adaptive security scheme algorithm.

57

excluded. For the rest of the security modes, first the ECDF is determined and then the selection follows based on the ECDF metric. In the second case, the security modes that do not meet the energy/time requirements are excluded. For the rest of the security modes, first the ECDF is determined and then the selection follows based on the ECDF metric. The proposed adaptive security scheme and its operation are described in the adaptive security algorithm, shown in Figure 4.2.

As mentioned earlier, the suggested adaptive security scheme provides two options for achieving the desired encryption strength at the lowest energy cost:

 For given security requirements for the requested service, the reliability function is used to return the most efficient option with respect to energy, for the specific security mode. This can be done by excluding the modes that do not meet the security requirements, and by ranking the modes after the elimination based on the reliability or the ECDF. The higher the ECDF, the highest the probability of finishing the encryption on time.

 In the case of battery powered devices, for a given energy threshold that derives from the battery state, the reliability function is used to return the most efficient option with respect to security, for the specific energy threshold. The modes that do not meet the time/energy requirements are excluded and the rest of the cases are ranked based on the ECDF/reliability and the selection is made based on this ranking.

Overall, the proposed adaptive security scheme consists of several security modes, each providing a different level of security, depending on the severity of the service requested. Each security mode operates using the appropriate security algorithms

58

and/or primitives. As the energy cost depends on the encryption parameters, each policy will induce a different level of energy consumption.

Using the empirical CDF, a probability metric is calculated for a specified energy threshold. In this way, one can either accept or reject the combinations according to the desired level of the probability, and, depending on whether they satisfy the requirements or not, a decision will be made, which implies that the combinations that do not meet the given constraints will be eliminated.

A general rule applied to most of the cases is that the highest probability of completing the encryption procedure prior to the time threshold will be selected, meaning that, for the specified threshold, the system will accomplish complete encryption in the most secure mode possible as well as at the lowest energy cost. Depending on the desired reliability, the most secure option will be selected.

In the case of a battery-level threshold, the security modes that do not meet the energy constraints are excluded and the rest of them are ranked according to their reliability/ECDF. As mentioned earlier, reliability is the probability that the system will continue the encryption process even after the given energy threshold. Therefore, the lower the reliability is, the higher the probability that the system will have finished the encryption procedure before the battery dies. Thus, the probability will be used to return the most efficient option with respect to security, for the specific energy threshold. In the case of specific security requirements, the probability will be used to return the most efficient option with respect to energy, for the specific security. Thus, the system will select the lowest reliability metric from the available options that meet the requirements of the desired security.

59