Research on Key Technologies of Quantum-Safe Metro-Optimized Optical Transport Networks

Abstract

This research introduces a novel physical-layer encryption technique for metropolitan-optimized optical transport networks (M-OTNs) that integrates real-time optical signal time-domain scrambling/descrambling with decoy-state quantum key distribution (DS-QKD). The method processes real-time optical data from the optical service unit (OSU) using a series of tunable Fabry–Perot cavities (FPCs), synchronized and updated with a running key. Experimental validation demonstrates secure communication within the optical network’s physical layer during standard OTU2 data transmission (10.709 Gbps), achieving an online transmission distance exceeding 100 km over typical single-mode fiber with a power loss of approximately 1.77 dB. The results indicate that this integrated approach significantly enhances the security of the optical physical layer in M-OTNs.

1. Introduction

In China, vast amounts of data are transmitted via metropolitan-optimized optical transport networks (M-OTNs), coinciding with a surge in network security threats. Traditional public key cryptography provides a security barrier against eavesdropping but is vulnerable to “collect first, decrypt later” attacks facilitated by quantum computers. Quantum key distribution (QKD) technology, which transmits single photons between communicating parties, ensures secure key agreement and detects potential eavesdropping attempts, thereby protecting data integrity [1,2,3]. Integrating QKD into M-OTNs can mitigate these security risks.

China Telecom has developed M-OTNs based on optical service units (OSUs), enabling flexible transport for various government and service applications through the adaptive mapping of optical service units. OSUs have been introduced to address the issues of low bandwidth utilization and insufficient service flexibility in traditional OTNs within metropolitan networks [4]. Additionally, an OSU-OTN architecture and service model have been designed specifically for government and enterprise dedicated networks, ensuring high security, reliability, and multi-service support [5]. An OSU routing optimization algorithm has also been developed to mitigate traffic load imbalance in optical inter-satellite links (OISLs) [6]. However, M-OTNs face challenges such as optical network eavesdropping and weak signal interception, particularly due to the increasing risk of eavesdropping in elastic optical networks, which results from flexible spectrum allocation [7], full-stack security threats, and hierarchical protection requirements [8]. Furthermore, a non-invasive traffic monitoring method using a temporary optical coupler based on fiber bending has been proposed for security analysis [9]. The advent of quantum computing poses significant threats to classical cryptographic systems, necessitating quantum-enhanced security paradigms. Current quantum communication solutions include continuous-variable QKD (CV-QKD), discrete-variable QKD (DV-QKD), and quantum secure direct communication (QSDC) [10,11,12]. These methods address vulnerabilities exposed by quantum computational attacks through different physical encoding mechanisms. However, QSDC operates at the Kbps level, while DV-QKD and CV-QKD achieve Mbps-level communication rates. The authors of [13] proposed teleportation principles in classical optical networks, demonstrating simultaneous quantum state transmission and 400 Gbps C-band data traffic over 30.2 km of optical cable. Additionally, a nonlinear all-optical encryption technique utilizing multi-photon absorption in carbon nanotubes has been proposed. This technology achieves high-security physical-layer encryption through dynamic optical intensity threshold adjustment, although it faces challenges in material and system integration [14]. The integration of QKD with transport layer security (TLS) 1.3 has been introduced to enhance optical network security against quantum threats [15].

To meet the communication rate requirements of 10 Mbps to 10 Gbps in M-OTNs, this paper proposes and experimentally validates a method combining the real-time optical signal time-domain scrambling/descrambling of OSU optical signals with decoy-state quantum key distribution (DS-QKD). The approach uses tunable Fabry–Perot cavities (FPCs) to process OSU optical signals in real time, synchronized with a running key. The DS-QKD system employs the decoy-state BB84 protocol and polarization coding to distribute seed keys. Experimental results confirm secure end-to-end OTU2 (10.709 Gbps) data transmission in access-side M-OTNs, with more than 100 km of optical data transmission over conventional single-mode fiber and a power loss of only 1.77 dB. This method offers an effective and adaptable solution for high-speed optical signal scrambling/descrambling.

2. Methods

2.1. OSU Service Mapping/Multiplexing

The packet service architecture utilizes a standardized 2.6 Mbps OSU framework via adaptive mapping and multiplexing, facilitating dynamic network slicing and multi-service convergence [16]. As illustrated in Figure 1, the ITU-T G.709-compliant solution introduces a containerization mechanism within the optical channel payload unit (OPU) structure. This layered abstraction enables (1) physical-layer isolation through time-frequency resource partitioning; (2) deterministic latency via fixed-frame scheduling; (3) cryptographic security by employing optical channel scrambling; and (4) bandwidth elasticity, supporting adaptive adjustment from 10 Mbps to 100 Gbps.

The primary requirement for OSU-hosted packet traffic is MAC-transparent transmission. The maximum bit rate of an OSU corresponds exclusively to the nominal Ethernet MAC bit rate, representing the highest assured bandwidth that an OSU can provide. An OSU is suitable for transmitting Ethernet traffic without a predetermined bit rate. The bit rate of an OSU varies with traffic fluctuations, specifically within the range of 4 to C times the PB benchmark bit rate ($R_{\mathrm{PB}}$). $C\times R_{\mathrm{PB}}$ denotes the maximum OSU bit rate, where C represents the maximum guaranteed bandwidth allocated by the OSU connection for Ethernet services.

Table 1. OSU bit rate for various Ethernet service rates.

Ethernet Service Rate	C Value	1 ${\mathit{R}}_{\mathbf{OSU}-max}$ (Kbps)	2 ${\mathit{R}}_{\mathbf{OSU}\_\mathbf{PLD}-max}$ (Kbps)
10 Mbps	4	10,400	10,020.833
100 Mbps	40	104,000	100,208.333
1 Gbps	400	1,040,000	1,002,083.333
10 Gbps	3996	10,381,800	10,003,296.875

¹ $R_{\mathrm{OSU}}=C\times R_{\mathrm{PB}},R_{\mathrm{PB}}=2.6$ Mbps. ² $R_{\mathrm{OSU}\_\mathrm{PLD}}=C\times R_{\mathrm{PB}}\times {\displaystyle \frac{185}{192}},R_{\mathrm{PB}}=2.6$ Mbps.

presents calculations of the maximum OSU bit rate for standard Ethernet services.

The ODU frame structure is divided into multiple payload blocks (PBs), with each OSU filling one or more PBs. The OSU bit rate is influenced by the type of packet traffic transmitted. The client signal is first mapped to the OSU and then multiplexed into the OPU frame through the optical service branch unit (OSTU) structure. The OSTU architecture organizes C PBs within an operational period spanning P consecutive OPU PB intervals. According to ITU-T G.709 [17], the peak capacity parameter C(measured in PBs) governs the theoretical maximum OSU throughput allocable per client port, which is given by

$$C={\displaystyle \frac{R_{\mathrm{client}}\times \left(1+R_{\mathrm{client}\_\mathrm{tolerance}}\right)\times \frac{192}{185}}{R_{\mathrm{PB}}\times \left(1-R_{\mathrm{PB}\_\mathrm{tolerance}}\right)}}$$

(1)

where $R_{\mathrm{client}}$, $R_{\mathrm{client}\_\mathrm{tolerance}}$, $R_{\mathrm{PB}}$, and $R_{\mathrm{PB}\_\mathrm{tolerance}}$ denote the client bit rate, client bit rate tolerance, OSU PB reference bit rate, and OSU PB reference bit rate tolerance (±20 ppm), respectively. The OSU maximum bit rate is calculated as follows:

$$R_{\mathrm{OSU}}=C\times R_{\mathrm{PB}}$$

(2)

The OSU maximum payload bit rate is calculated as follows:

$$R_{\mathrm{OSU}\_\mathrm{PLD}}=C\times R_{\mathrm{PB}}\times {\displaystyle \frac{185}{192}}$$

(3)

The OSU adopts the OPU division method based on the PBs, as shown in Figure 2. For a given OPU, each transmission cycle consists of P consecutive PBs, and the PB size of each OPU is 192 bytes.

Table 2. Corresponding p-values and PB values for different OPUks.

OPU Type	${\mathit{R}}_{\mathbf{OPU}\_\mathbf{PLD}}$ (Kbps)	p Values	PB Actual Bit Rate (Mbps)
OPU0	1,238,954.310	476	2.602845189
OPU1	2,488,320.000	956	2.602845188
OPU2	9,995,276.962	3840	2.602936709
OPU3	40,150,519.322	15,426	2.602782272
OPU4	104,355,975.330	40,096	2.602653016

presents the corresponding p values of ODUs with different rates under the condition of 2.6 Mbps as the OSU PB benchmark bandwidth ($R_{\mathrm{PB}}$), the actual PB bit rate of different OPUks is close to the OSU PB benchmark bandwidth, and the p values can be calculated using Equation (4):

$$p={\displaystyle \frac{R_{\mathrm{OPU}\_\mathrm{PLD}}\times \left(1-OPU\_PLD\_BR\_tolerance\right)}{R_{\mathrm{PB}}\times \left(1+1000\mathrm{ppm}\right)}}$$

(4)

The mapping and multiplexing of OSU services to OPUs are implemented based on transmission cycles. Each transmission cycle sequentially indexes p payload blocks (PBs) as $P{B}_1$ to $P{B}_p$. During the mapping of OSU services to OPUs, a maximum of C PBs is allocated per service per cycle. For m multiplexed OSU services, their respective PB allocations $C_1,C_2,\dots ,C_m$ must satisfy ${\sum }_{i=1}^m{C}_i\le p$, ensuring that the OPU provides sufficient bandwidth to accommodate all services. To guarantee bandwidth, a maximum of C PBs should be allocated to each OSU in each transmission cycle during the mapping and multiplexing of OSU traffic to OPUs.

The mapping and multiplexing mechanism employs a method analogous to sigma-delta modulation: C out of p PBs are evenly distributed to a given OSU service in each transmission cycle, as illustrated in Figure 3. The actual number of PBs used by each OSU service depends on its traffic volume, and any vacant PBs can be assigned to the OSU IDLE frame.

Consequently, OSUs implement hard-pipeline isolation, a stringent resource isolation mechanism that ensures the exclusive transmission of service flows through fixed resource reservation (PBs) within the OPU. This approach eliminates the need for services to compete for resources, thereby reducing transmission jitter.

2.2. Time-Domain Scrambling/Descrambling Combined with DS-QKD

The operational mechanism of the real-time OSU optical signal time-domain scrambling system integrated with the DS-QKD system is illustrated in Figure 4. This system establishes a symmetric key-based scrambling/descrambling framework, involving both the DS-QKD transmitter and receiver. The key exchange occurs via DS-QKD, facilitating periodic and random renewal of the shared seed key between two authorized users, which are known exclusively to the transmitter and receiver. On the transmitting side, the DS-QKD transmitter sends quantum-generated random keys over the quantum channel to the receiver without transmitting any accompanying message data. The scrambler employs the running key, generated by field-programmable gate arrays (FPGAs), to scramble the OSU optical signal in real time. Given that the M-OTN OTU2 interface’s line rate is approximately 10.709 Gbps, the FPGAs used for scrambling and descrambling are XILINX 7K325T devices, which support a maximum SerDes (GTX) line rate of 12.0 Gbps. On the receiving side, the DS-QKD receiver captures the quantum bits from the transmitter and derives the seed key through the protocol channel. The descrambler then applies the identical running key used by the transmitter to restore the scrambled OSU optical signal. Synchronization between the transmitter and receiver for scrambling and descrambling is maintained via the FPGA synchronization channel.

It should be noted that all communications within the protocol, synchronization, and data channels consist of conventional signals, allowing these channels to be multiplexed using a wavelength-division multiplexer (WDM) and transmitted over the classical communication channel. Typically, at the start of transmission, the DS-QKD transmitter and receiver exchange and store a quantum key. The FPGA within the transmitter generates the running key based on the seed key and synchronization marker. The OSU optical signal entering the scrambler undergoes real-time scrambling by dynamically adjusting physical parameters according to the running key. Simultaneously, the synchronization marker, transmitted via the synchronization channel, ensures the alignment of the scrambling and descrambling initiation signals. At the receiver, the same running key derived from the seed key is used to demodulate the received signal, triggered by the synchronization marker. This dynamic key approach significantly enhances security, as an eavesdropper must not only decipher the ultra-high-speed scrambled signal but also intercept the key before the next update cycle.

Fiber Bragg Grating (FBG) and FPCs are two common optical resonant structures. FBG is sensitive to temperature and strain, while FPCs are sensitive to cavity length changes. Given the difficulty of controlling the temperature of FBG, FPCs are modularized to control cavity length more effectively. The time-domain scrambling of the OSU optical signal is achieved through FPCs, as illustrated in Figure 5a. When an OSU optical signal enters a single-fiber delayer, it undergoes multiple reflections between the dual reflective surfaces of the cavity (Figure 5b). Each scrambling/descrambling controller comprises 12 FPCs. While increasing the number of FPCs enhances the scrambling effect, it also introduces significant dispersion, necessitating adjustment using dispersion compensation fiber (DCF). This reflection mechanism splits the collimated light into multiple sub-beams through the cavity structure, with each beam acquiring distinct transmission delays. The delayed sub-beams are subsequently focused through a lens assembly into the output fiber. The temporal characteristics of the composite output signal are determined by the coherent superposition of these differentially delayed optical components.

The scrambling implementation employs independent temperature-controlled modules (TCMs) for each fiber delayer, enabling the dynamic adjustment of cavity parameters to achieve temporal-domain signal scrambling. The FPCs have been modularized, with a working temperature ranging from 0 °C to 50 °C. For subsequent experiments, the ambient temperature is maintained at room temperature, specifically within the range of 20 °C to 25 °C.

This scrambling mechanism induces waveform reconstruction through two primary effects:

• Temporal rearrangement of optical pulses, creating new waveform configurations;
• Controlled inter-symbol interference through adjacent-bit overlap.

The resultant signal mixture exhibits enhanced resistance to eavesdropping attempts due to the following:

• Non-linear cross-interaction between overlapping bits;
• Noise-equivalent signal degradation from timing jitter;
• Irreversible scrambling transformations.

The above scrambling process can be mathematically modeled as a convolution operation between the input optical signal $z_{\mathrm{osu}}\left(t\right)$ and the FPC’s impulse response $h\left(t\right)$ as follows:

$$z_{osu}^{\prime }\left(t\right)={\int }_{-\infty }^{\infty }{z}_{\mathrm{osu}}\left(\tau \right)h(t-\tau )d\tau =z_{\mathrm{osu}}\left(t\right)\otimes h\left(t\right)$$

(5)

where $z_{\mathrm{osu}}^{\prime }\left(t\right)$ is the time-scrambled signal and “⊗” is the convolution operation. $z_{\mathrm{osu}}\left(t\right)$ is the input OSU optical signal. $h\left(t\right)$ is the FPC impulse response, and its inverse Fourier transform is expressed as

$$h\left(t\right)={\int }_{-\infty }^{\infty }r\left(\omega \right)exp(-i\omega t)d\omega$$

(6)

where the frequency response of FPC $r\left(\omega \right)$ is denoted as

$$r\left(\omega \right)={\displaystyle \frac{r_1-r_{2}exp\left(i2\omega ndcos{\theta }_0\right)}{1-r_1{r}_{2}exp\left(i2\omega ndcos{\theta }_0\right)}}$$

(7)

Let $r_1$ and $r_2$ be the reflectance of the two surfaces of the FPC, respectively. The optical frequency, denoted as $\omega$, is defined by the equation $\omega =2\pi /\lambda$, where $\lambda$ represents the wavelength of light in a vacuum, n denotes the refractive index of the FPC substrate, and d denotes the thickness of the FPC substrate. The angle of incidence, ${\theta }_0$, is the angle at which the optical signal enters the FPC.

It can be seen from Equations (6) and (7) that the impulse response of the FPC $h\left(t\right)$ is correlated with $r_1$, $r_2$, n, and d. In subsequent experiments, we ensured that ${\theta }_0$ was 0° and the incident polarization had no influence on the experimental data. To achieve time-domain randomization of the OSU optical signal, a TCM can be used to adjust the optical thickness of the FPC $nd$.

The OSU optical signal after N-level cascaded FPC scrambling is denoted as

$$z_{\mathrm{osu}}^{\prime }\left(t\right)=z_{\mathrm{osu}}\left(t\right)\otimes h_1\left(t\right)\otimes h_2\left(t\right)\otimes \dots \otimes h_N\left(t\right)$$

(8)

The OSU optical signal reflected from the FPC array of the scrambler is denoted as

$$\begin{array}{c}{z}_{\mathrm{osu}}^{\prime \prime }\left(t\right)=z_{\mathrm{osu}}\left(t\right)\otimes h_1\left(t\right)\otimes h_2\left(t\right)\otimes \dots \otimes h_N\left(t\right)\otimes h_1^{\prime }\left(t\right)\otimes h_2^{\prime }\left(t\right)\otimes \dots \otimes h_N^{\prime }\left(t\right)\\ =z_{\mathrm{osu}}\left(t\right)\otimes H\left(t\right)\otimes H^{\prime }\left(t\right)\end{array}$$

(9)

where $H\left(t\right)=h_1\left(t\right)\otimes h_2\left(t\right)\otimes \dots \otimes h_N\left(t\right)$ is the impulse response of the scrambler and $H^{\prime }\left(t\right)=h_1^{\prime }\left(t\right)\otimes h_2^{\prime }\left(t\right)\otimes \dots \otimes h_N^{\prime }\left(t\right)$ is the impulse response of the descrambler. The solution to time scrambling signal interference follows a principle similar to that in [18], which is based on the thermal light effect and uses a dynamic reconfigurable code synchronization solution as a key approach. Let $H^{\prime }\left(t\right)=H^{*}(-t)$. It follows from Equation (9) that

$$z_{\mathrm{osu}}^{\prime \prime }=z_{\mathrm{osu}}\left(t\right)\otimes H\left(t\right)\otimes H^{*}(-t)=z_{\mathrm{osu}}\left(t\right)\otimes \left[H\left(t\right)\otimes H^{*}(-t)\right]$$

(10)

The superscript “*” represents the complex conjugate, whereas $H\left(t\right)\otimes H^{*}(-t)$ indicates the auto-correlation function of the scrambling and descrambling impulse responses. When this function is equivalent to the Dirac delta function ($\delta \left(t\right)=H\left(t\right)\otimes H^{*}(-t)$), the descrambler’s output accurately reproduces the input of the corresponding scrambler. This allows the authorized receiver to completely reconstruct the original OSU optical signal using the correct synchronization key. If the descrambler’s impulse response does not match the complex conjugate of the scrambler’s impulse response, i.e., $H^{\prime }\left(t\right)\ne H^{*}\left(t\right)$, the descrambler’s output becomes the cross-correlation function of two distinct FPC arrays. Consequently, the OSU optical signal undergoes double scrambling in the time domain, making it impossible for an eavesdropper to descramble the signal without the correct running key.

3. Results

To validate the proposed method, an integrated system combining time perturbation with DS-QKD to achieve OSU optical physical-layer security has been developed, as illustrated in Figure 6. The manufacturers and models are listed in

Table 3. Device manufacturers and models.

Device	Manufacturer	Model
M-OTN	CETC, Guilin, China	Design with XILINX 7V690T
Scrambler/descrambler	CETC, Guilin, China	Design with XILINX 7K325T
DS-QKD	QuantumCTek, Hefei, China	QKD-POL1250-S

. This DS-QKD system employs the decoy-state BB84 protocol using polarization coding to generate a seed key via the quantum and protocol channels. The design considerations stem from the fact that these systems are highly sensitive to degradation and cannot amplify single photons, which necessitates a low-temperature environment. Consequently, a single G.652 standard single-mode fiber (SMF) is utilized as the quantum channel, spanning a total length of 100 km without any optical parametric amplifiers. The quantum channel exhibits linear loss (attenuation) of 10 dB over 100 km.

The secure seed key is fed into the FPGA at both the transmitter and receiver ends. The FPGA then generates a running key to control the scrambling and descrambling processes. Key synchronization between the scrambler and descrambler is achieved through a dedicated synchronization channel. Subsequently, the M-OTN device’s output OSU optical signal is scrambled using an FPC array. Moreover, the decoy BB84 protocol signal for key generation, the synchronization signal for key alignment, and the scrambled optical signal are multiplexed with the classical channel via a WDM and transmitted over the same 100 km single-mode fiber at wavelengths of 1546.92 nm, 1548.51 nm, and 1550.12 nm, respectively.

In the classical channel, when the transmission distances are 50 km and 100 km, the optical signal losses are approximately 10 dB and 20 dB, respectively. To mitigate power loss due to scrambling/descrambling and fiber transmission, two erbium-doped fiber amplifiers (EDFAs) are employed. The noise superimposed on the optical signal by the EDFAs is minimal and can be neglected. Additionally, DCF is used to address dispersion issues over the 100 km single-mode fiber. After scrambling/descrambling, the bit error rate (BER) and eye diagram can be measured using an Ethernet tester (MTS 5800, VIAVI Solutions Inc., Chandler, AZ, USA) and an optical oscilloscope (TEKTRONIX DSA8200, TEKTRONIX Inc., Beaverton, OR, USA), respectively.

The eye diagram of the OSU electrical signal, as shown in Figure 7, is open and clear. The experimental results of OSU optical signal scrambling and descrambling are depicted in Figure 8. Figure 8a,b show the unscrambled and scrambled signals, respectively. Due to the use of an FPC array for scrambling the overlapping bits, the scrambled signal is significantly different from the original 10.709 Gbps non-return-to-zero (NRZ) data-modulated optical signal. This indicates that the scrambler completely disrupted the temporal position relationship between different bits, rendering the OSU optical signal noisy and not digitizable. The eye diagrams after descrambling, as shown in Figure 8c, reveal an open eye diagram when using matched key descrambling. Conversely, when using unmatched key descrambling, the eye diagram remains closed, and the outline does not exhibit any numeric features of “0” or “1”, as depicted in Figure 8d. Therefore, an eavesdropper cannot simultaneously acquire the scrambling and descrambling strategies, as well as the synchronization key, making it impossible to extract OSU data via optoelectronic transformation.

The performance of the running key rates in the DS-QKD system is illustrated in Figure 9a, where data were collected at six different transmission distances: 0, 20, 40, 60, 80, and 100 km. Specifically, the secret key rates were 14,125 bps, 12,511 bps, 7962 bps, 4468 bps, 2679 bps, and 797 bps, respectively. Additionally, the bit error rate (BER) performance in the back-to-back (B2B) scenario is presented in Figure 9b. The BER and average received optical power (${\mathrm{P}}_{\mathrm{R}}$) were measured over a transmission distance of 100 km. These measurements were conducted at a full throughput of 10.709 Gbps at the network-to-network interface (NNI), achieving error-free operation (with the bit error rate satisfying forward error correction (FEC) requirements).

For the B2B scenario, Figure 9b shows the results with and without scrambling using dashed lines. Compared to the unscrambled OSU optical signal, the introduction of scrambling triggered by QS-QKD led to a slight degradation in BER performance. At $BER={10}^{-9}$, the ${\mathrm{P}}_{\mathrm{R}}$ values for the unscrambled and scrambled OSU optical signals were approximately −21.31 dBm and −20.40 dBm, respectively, resulting in an optical power loss of about 0.91 dB for the scrambled signal.

After 100 km of transmission, as indicated by the solid lines in Figure 9b, similar BER performance was observed. At $BER={10}^{-9}$, the ${\mathrm{P}}_{\mathrm{R}}$ values for the unscrambled and scrambled OSU optical signals were approximately −19.64 dBm and −17.87 dBm, respectively, leading to an optical power loss of about 1.77 dB for the scrambled signal.

These findings demonstrate a real-time optical signal temporal scrambling/descrambling method, which enhances the physical-layer security of OSUs in M-OTNs. Cryptographic keys are exchanged between transceivers and receivers via DS-QKD over a 100 km range. An eavesdropper without the scrambling strategy and synchronization key cannot correctly extract the digital features of the OSU optical signal, rendering it indistinguishable from noise and making interception and subsequent analysis nearly impossible. Synchronous scrambling and descrambling using the correct quantum key can achieve optical physical-layer security, albeit with a slight attenuation in optical power.

4. Conclusions

This paper proposes and experimentally demonstrates a real-time optical signal temporal scrambling/descrambling method to enhance physical-layer security in M-OTNs. In this method, real-time OSU optical signal processing is performed using a tunable FPC array with a dynamically updated and synchronized quantum key. The DS-QKD system implements the decoy-state BB84 protocol with polarization encoding to provide seed keys. Compared to traditional electrical-domain scrambling/descrambling methods, the proposed method introduces an additional and distinct security layer into the M-OTN system, thereby providing robust security protection at the optical physical layer.