1. Introduction

Quantum key distribution (QKD) systems developed to securely share security keys between remote parties, denoted as Alice and Bob, can be divided into fiber-based QKD and free-space QKD depending on the channels used [1–13]. Although QKD research based on fiber is more mature in terms of commercialization, free-space QKD research using satellites, aircrafts, and balloons has been actively studied because it could potentially solve the transmission distance problem [14–26]. As with fiber-based QKD, risks of quantum hacking exist because of the incompleteness of the free-space QKD system hardware. Various kinds of attacks (e.g., blinding of avalanche photon diodes and Trojan-horse attacks) are also applicable in the free-space QKD system, although they have originally been studied in fiber-based QKD [27–29]. Many reports also detail the risks of hacking methods applicable only to free-space QKD systems [30–32]. Unlike fiber-based QKD system attacks, which focus on detector attacks, free-space QKD system hacking studies typically consider the imperfections of the quantum light source as major attack paths [27–29, 33–43]. An eavesdropper can detect marginal characteristic differences of the photons from the four laser diodes (LDs) used in free-space QKD systems, such as the spatial, spectral, and temporal modes of the photons. The eavesdropper can infer the quantum state of the photon from these detected results [32]. Although these methods successfully demonstrate the risk of quantum hacking, they are easily prevented by methods to make the four LDs indistinguishable, such as the temperature control of the diodes.

However, we discover a risk of quantum hacking even when four LDs with the same characteristics are used. An external laser can change the spectral characteristics of the LDs, and the degree of wavelength variation of a particular LD can be controlled depending on the amount of light irradiated. In this study, we propose a new free-space QKD system hacking method based on this phenomenon of the wavelength control using an external laser. We explain the concept of the proposed hacking method in Chapter 2. The proof-of-principle experimental results and summary of our work, including possible countermeasures, are provided in the following chapters.

2. Concept of quantum hacking through laser diode side-channel

Free-space QKD usually utilizes polarization states to encode the bit information. There are two ways to generate the quantum information: one is to modulate the polarization state from one LD, while the other is to use four LDs which generate different polarization states. Although the former has an advantage in security, the latter method is more commonly used because of its better high-speed operation characteristics. The drawback of the second method is that there can be discrepancy the characteristics of each of the four LDs, such as the spatial, spectral, and temporal modes. Consequently, Eve (the eavesdropper) can leak quantum key information and threaten system security using these differences [32]. To prevent this, Alice’s four LDs are controlled in characteristics through temperature management and mode cleaning, ensuring that they have the same characteristics and cannot be distinguished by Eve. However, we have shown that one can easily distinguish each LD by changing the wavelength of the individual LDs using a strong external laser input. Using the scheme shown in Fig. 1, Eve can determine all of Alice’s polarization information while simultaneously pretending to be Alice and participating in quantum communication with Bob.

 

Fig. 1 Concept of quantum hacking by manipulating laser diodes in Alice’s system.

Download Full Size | PDF

Eve is located between Alice and Bob in our proposed scheme. In general, each LD in Alice’s system is preceded by a polarizer (P), which allows the different LDs to have different polarization states. For example, LDs 1, 2, 3, and 4 provide diagonal (D), anti-diagonal (A), horizontal (H), and vertical (V) polarizations, respectively. Alice randomly selects only one of the four LDs for operation, thereby creating a laser source with a specific polarization state. Alice sends radiation from this selected LD to Bob. Bob then uses a single-photon detector to measure the polarization state. Alice must equalize the LD wavelength during the QKD communication. If the wavelength of the specific LD varies, Eve can know which LD is working, and hence, she can easily steal Alice’s polarization information. In general, Alice uses a thermoelectric cooler (TEC) to change the LDs’ temperature and make the wavelength equal.

However, we have discovered a side-channel: the internal temperature of Alice’s LD can be changed by external influences. In other words, it is possible to freely adjust Alice’s LD wavelength. If Eve can adjust the four LD wavelengths by different amounts, Eve can obtain Alice’s polarization information and simultaneously send a signal to Bob, pretending to be Alice. Figure 1 shows that Eve’s system consists of Eve’s LD (Eve_LD), a wavelength division multiplexer (WDM), polarization beam splitter (PBS), half wavelength plate (HWP), single-photon detector (SPD), and LDs of the same model as Alice’s. Eve uses the following procedure when hacking free-space QKD systems: first, Eve creates a strong laser source with a certain polarization (V) and sends the light to Alice. This strong laser radiation is either reflected or passed through the polarizers (P) and enters the LDs of Alice’s system. The amount of reflected or transmitted radiation differs depending on the polarization setting of P. For example, all laser radiation power on LD4 is transmitted when a strong laser radiation controlled by V-polarized light is incident on a P fixed with V-polarized light. However, the same radiation entering P fixed with an H-polarized light cannot be transmitted to LD3. The laser radiation for P with D and A polarizations can be transmitted or reflected through LD1 and LD2 with 50% intensity. As a result, LD1, LD3, and LD4 receive different amounts of laser irradiation, such that the temperature of each LD changes by a different amount, thereby simultaneously changing the wavelength differently in each LD. LD1 and LD2, which receive the same amount of laser light, experience the same change in wavelength. With this phenomenon, Eve can classify the diodes by wavelength using WDM. For example, the wavelength of LD4 is located in λ3 of the largest variation wavelength because the LD4 temperature is dramatically increased by receiving 100% of the external laser source’s incident power. Meanwhile, the LD3 maintains its previous wavelength (λ1) because the P before LD3 reflects all the strong laser power. The LD1 and LD2 wavelengths are located together in λ2, because the LDs obtain the same amount of 50% energy from the strong laser source. Eve can easily separate the D and A polarizations in λ2 if she uses a HWP and a PBS. The outputs of the WDM and PBS connect with the SPD to determine the polarization generated by Alice. By this method, Eve can determine all polarization information from Alice’s system and simultaneously send another single photon to Bob with the same polarization information as that from Alice using LDs 1-4. Therefore, Bob cannot recognize the Eve’s existence. In a sifting process, Eve can generate full quantum keys equal to those of Alice and Bob without increasing the quantum bit error rate if she hacks into a classical channel between Alice and Bob.

3. Experiments and results

3.1. Experiments

This new method of quantum hacking was proven by first measuring how the wavelength of Alice’s LD is changed by the laser source generated by Eve’s LD. We used the experimental setup shown in Figs. 2(a) and (b) for the proof-of-principle experiment. Figure 2(a) shows the setup with a strong laser source (0-600 mW) of a femtosecond laser (FSL) to measure the wavelength change of Alice’s LD. The wavelength change of the laser source was measured using a spectrometer (SP). Fig. 2(b) shows the experimental setup to investigate the change of the wavelength using only a laser power of 0-25 mW applied by a general LD of 492 nm. At this time, a Fabry-Perot interferometer (FPI)was utilized to accurately measure the wavelength of Alice’s LD. In these experiments, each LD in Alice’s system may not have the same wavelength. The difference was caused by the different inherent characteristics of Alice’s LDs. The unequal wavelengths of the LDs must be resolved before a proof-of-principle experiment of the proposed hacking scheme is executed. We adjusted the amount of laser driving current and the temperature with a TEC controller (TECC) to solve this problem. We equalized the LD wavelength in the experiments in two different ways. First, only the driving current of the LDs was adjusted to achieve the single output wavelength of 808 nm. Second, the temperature was controlled with the same driving current of 46 mA to Alice’s LDs. The internal temperature of the LD was adjusted to produce the laser output wavelength of 808 nm using TECC.

 

Fig. 2 Experimental setup of quantum hacking using external strong laser pulses. (a) Quantum hacking experiment with femtosecond laser and spectrometer. (b) Quantum hacking experiment with Fabry-Perot interferometer. Alice LD: Alice’s laser diode; TECC: thermoelectric cooler controller; DM: dichroic mirror; NDF: neutral-density filter; SP: spectrometer; Eve LD: Eve’s laser diode; FSL: femtosecond laser; FPI: Fabry-Perot interferometer; FPIC: Fabry-Perot interferometer controller; and OSC: oscilloscope.

Download Full Size | PDF

The 700 nm (or 492 nm) output of Eve’s laser was reflected by a dichroic mirror (DM) and incident on Alice’s LD, such that the output of Eve’s LD be incident on the same path as that of Alice’s LD. The laser radiation generated by Alice’s LD passed through the DM, but the 700 nm (or 492 nm) light from Eve was blocked. A neutral-density filter (NDF) was added in Fig. 2(a) to attenuate the laser radiation, thereby preventing the SP from encountering a strong laser radiation. Figure 2(b) shows that the FPI controller (FPIC) controlled the amplitude of the FPI output and adjusted the oscilloscope (OSC) for measurement.

3.2. Results

Figure 3(a) shows the measurement result of the wavelength change according to the temperature change of Alice’s LD. The temperature of Alice’s LD was varied from 20 to 30 C using the TECC within Alice’s system, thereby confirming that the wavelength of LDs 1-4 gradually increased from 808 to 811 nm with the increasing temperature. The wavelength change can be manipulated as desired if we introduce laser radiation that generates sufficient thermal energy in Alice’s LDs. Figure 3(b) presents the result of the incident strong laser radiation with 0 to 600 mW power using an external FSL focused on LD 1. The wavelength of Alice’s LD was changed from 808 nm to 814 nm by the strong laser radiation. This result shows that the laser radiation incident from outside generates internal heat energy.

 

Fig. 3 (a) Measurement of wavelength change with increasing internal temperature of Alice’s laser diode. (b) Wavelength change of Alice’s LD according to the laser power generated by Eve’s LD.

Download Full Size | PDF

Figure 4 shows more details on the measurement of the change in wavelength when laser radiation was applied to Alice’s LD using the FSL. The results were classified into two cases of with or without the temperature control using a TECC [Fig. 2(a)]. The wavelengths of LDs 1-4 increased from 808 nm to a minimum of 813 nm and a maximum of 816 nm when the laser power of Eve’s LD increased from 0 to 600 mW. The amount of increase of each LD wavelength differed, but the data confirmed that the wavelengths of the LDs can be dramatically changed by 5 to 8 nm due to the external laser radiation. Thus, Eve can distinguish among the LDs of Alice’s system based on different laser wavelength if she generates a strong laser radiation with vertically polarized light. As shown in the above experiment, Eve uses a strong laser radiation to generate many wavelength variations and distinguish each LD.

 

Fig. 4 Wavelength of the laser diodes versus the incident laser power generated by the femtosecond laser.

Download Full Size | PDF

Meanwhile, even a weaker light source can distinguish each LD because measuring the wavelength more precisely was possible with an FPI instead of a SP. Figure 5 shows the results of the precise measurement experiment using an FPI [Fig. 2(b)] by changing the wavelength of Alice’s LD using a general 492 nm LD. The wavelengths of LDs 1-4 were changed by 0–80 pm when temperature control was not performed using TECC while the power of the 492 nm LD was changed from 0 to 25 mW. The amount of the wavelength change with the temperature control was measured as 0–40 pm. Without temperature control of the LD, the temperature of the LD directly increased according to the amount of laser radiation received. With temperature control, the temperature change was restricted, explaining smaller changes in wavelength in this case. However, the changes in the wavelength of Alice’s LDs by external laser radiation in both cases were identified, and distinguishing each LD was possible. Therefore, there is no problem in obtaining all of Alice’s polarization information in both cases. Without being noticed, Eve can observe the communication between Alice and Bob if she sends a signal to Bob using the obtained polarization information.

 

Fig. 5 Frequency change and wavelength change versus incident laser power measured by Fabry-Perot interferometer.

Download Full Size | PDF

4. Conclusion

In this study, we discovered a side-channel in free-space QKD. Alice’s LDs can experience changes in temperature by the reception of external laser radiation. Moreover, the wavelength of the LDs can be changed by this temperature manipulation. We used this side-channel to prove and verify a quantum hacking method that allowed Eve to acquire the complete polarization information of Alice without informing her presence. Therefore, Eve, pretending to be Alice, can transmit signals to Bob. The hacking method can be directly applied to practical free-space QKD systems. Such a quantum hacking method should be prevented as it can expose 100% of the polarization information of the free-space system. A narrow band-pass filter (BPF), which is normally used in a free-space QKD system, may be helpful in preventing Eve’s strong laser with different wavelengths. However, some reports [44, 45] described the damage of BPF by stronger lasers, which means that the proposed hacking method can threaten the security. In addition, the use of a laser with a wavelength similar to that of Alice’s LD could also hack the QKD system. The following protection methods must be implemented to prevent the discovered hacking method: first, an external strong laser monitoring unit can detect Eve’s attempt. Random variations in the LD wavelength under Alice’s control can be a second solution. The use of passive devices, such as isolators, would also be helpful. The proposed quantum hacking method could be completely prevented if studies on these methods are applied to free-space QKD. Our hacking scheme can be applied to a free-space QKD system using four laser diodes, but generally cannot be directly applied to a fiber-based QKD that is usually implemented with only a single laser diode. However, in the future, more studies would be required for secure systems due to the wavelength change caused by an external laser may generate side-channels in fiber-based QKD systems.