HTML5 Webook

40/84

detectors, as a window pane was set in the light path to the detector of Bob, this experiment simulated one of the typical scenarios of wiretapping in free space optical com-munication where Eve wiretaps using a more sensitive detecting system than that of Bob.Under the experimental conditions mentioned above, Alice transmitted a pseudorandom number of length 2×106 in a transmission duration of 200 ms. en, we derived the intensity distribution of the channel based on the output of photo-detectors of Bob and Eve and calculated mutual information, I(X ; Y) and I(X ; Z) for every 4 ms, respec-tively. Here, we assumed that, realistically, the surroundings of Bob were monitored and guarded rmly and another wiretapper who could obtain more information than Eve on the roof could not exist. Under these conditions, we evaluated the transmission performance of the wiretap channel coding, which is the simplest physical layer cryp-tography. Although the performance of wiretap channel coding is measured by secrecy capacity as described above, we cannot estimate the secrecy capacity because we did not optimize power and the frequency of 0s and 1s in the pseudorandom number sequence in this study. Hence, we evaluate the amount of condentially transmissible infor-mation using the secrecy rate which is a simple dierence of the mutual information,);();(ZXIYXIRS .We performed such analysis in ve time zones, 14:43, 15:57, 16:33 (sunset time of the day), 17:37, and 18:10. In each time zone, ten transmissions are conducted by inter-vals of 20 seconds.e time variations of the secrecy rate calculated from the experimental data obtained at 16:43:20 and 17:37:00 are shown in Fig. 4. Here, we note that Bob’s channel was al-most error free in this experiment.e secrecy rate changed dramatically from 10 Mbps (from 24 ms to 28 ms) to 0 bps (from 144 ms to 148 ms) during the 200 ms period at 16:43:20, that was just aer sunset, as shown in Fig. 4(a). On the other hand, the time variation of the secrecy rate was small at 17:37:00, that was aer sunset, as shown in Fig. 4(b). From the above, it is shown that fatal information leakage occurred due to the eect of intensity modulation or beam wandering caused by atmospheric uctuation just before and aer sunset, however such eects are suppressed during night so that stable condential transmission would be realized using physical layer cryptography.In order to examine such atmospheric eects discussed above, we evaluated the probability that secrecy rate RS is smaller than a certain threshold Rth, )Pr()(SththoutRRRP.for the data obtained in the above ve experimental time zones. is is called secrecy outage probability Pout(Rth), because it can be deemed as a possibility of fault in con-dential transmission of designed code, if this threshold is regarded as a target rate determined at design of code. We calculated this Pout(Rth) for the ve time zones based on the data taken in the experiment and the results are shown in Fig. 5. It is impossible to decrease the secrecy outage prob-ability to 0 before sunset no matter how low the threshold Rth is set. On the other hand, there exists a threshold that FiF5Secrecy outage probability calculated from the experimental data obtained on 17th November 2015 [38].10M1M100k10k1k110-110-210-318:10 - 18:1316:33 - 16:3615:57 - 16:00Sunset time14:43 - 14:4617:37 - 17:40Threshold rate [bps]Secrecy outage probabilityFiF4(a) Secrecy rate calculated from experimental data obtained at (a) 16:34:20 and (b) 17:37:00, on November 17, 2015. Time interval of each measurement is 4 ms and corresponds to pseudorandom number sequences of length 4 × 104 [38].04080120160200Duration [ms]6M10MSecrecy rate [Mbps]2M04080120160200Duration [ms]1k10k100kSecrecy rate [bps]1M10M(b) 17:37:00(a) 16:34:203 Quantum Key Distribution Network36　　　Journal of the National Institute of Information and Communications Technology Vol. 64 No. 1 (2017)