1. Introduction

With the increasing demand on communication network security, physical layer security in optical networks has attracted a lot of research attentions lately. Compared with encryption schemes in the higher layers, security enhancement in the physical layer can provide additional protections to the network, which can be highly preferable in some applications, e.g. defense and financial enterprise networks.

Optical steganography transmission was recently proposed as a method of secure transmission over a public optical communication network [1–6]. In optical steganography transmission, an optical stealth channel is concealed underneath existing public channels such as wavelength division multiplexing (WDM) channels, in both spectral and temporal domains to achieve security by avoiding attention. It has been shown that the optical steganography transmission technologies can improve communication confidentiality [1,2]. In some of the demonstrated optical steganography transmissions, signal encoding and decoding technologies based on optical code-division-multiple-access (OCDMA) were employed [1,2,5,6]. With the OCDMA technology, optical signals are first destructed (temporally spread to noise-like signals) with an encoder before the transmission and then reconstructed with a matched decoder at the receiver side. Since the encoded OCDMA signal requires matched decoding at the receiver side, the adoption of OCDMA technologies for optical steganograghy transmission can provide enhanced security features [7] for optical stealth channels. In the meantime, optical steganography transmission technologies can support the potential deployment of OCDMA communications over existing public networks.

So far in all the theoretical analysis [1,2,6] and demonstrated experimental systems [5] of OCDMA-based optical steganography transmissions, OCDMA techniques based on 1D coherent spectral phase coding (SPC) [8] or 2D incoherent wavelength hopping time spreading (WHTS) [9] were used for the optical stealth signals. In the previous experimental demonstration of SPC-encoded optical stealth transmission over a public network [5], the public channel employs a short pulse laser source, which is not a common light source for most already deployed WDM-based optical networks. When WHTS encoding scheme [6] is used for optical stealth channels, the total number of hopping wavelengths is usually equal to or less than the number of public WDM channels, which can limit the autocorrelation peak intensity of the decoded signal, and consequently limits the performance of the stealth channel. Furthermore, the stealth signal using SPC or WHTS-OCDMA may be seriously degraded due to spectral distortion caused by the dynamical adding, dropping, or cross-connecting of wavelengths in the public WDM networks.

To overcome the above problems, here we use temporal phase coded optical stealth signals generated with super-structured fiber Bragg grating (SSFBG) [10–12], and demonstrate the transmission of optical stealth signals over a public WDM network. In our proposed scheme, the stealth signal is spread by a temporal phase encoder. The optical spectra of the noise-like stealth signals can be covered underneath the accumulated ASE noise spectrum of the public signals during transmission. By employing multiple notch filters to remove the interference from the public channels [13], error-free detection of the stealth channel can be achieved without using complicated optical thresholder [14–16].

2. Spectral notched temporal phase coding-OCDMA

In a temporal phase coding (TPC)-based OCDMA system, a series of coherent optical pulses are generated by a reflective temporal phase encoder with injection of a short optical pulse train, and these duplicated pulses carry the phase pattern determined by the OCDMA code. An autocorrelation (AC) peak is constructed along with some OCDMA side lobes (wings) after the encoded signal passing through a matched temporal phase decoder. The SSFBG used in our TPC-OCDMA system has a slowly varying refractive-index modulation profile. The phase pattern defined by an OCDMA code is realized by including phase shifts between different segments of the SSFBG [10]. The reflection spectrum of an SSFBG encoder can be calculated from its temporal characteristics [11].

We experimentally investigate the impact of notch filtering on SSFBG-encoded optical pulses in a TPC-OCDMA system. Figure 1 shows the schematic of our spectral notched TPC-OCDMA encoding and decoding setup. A 2.5Gb/s optical pulse train is injected to a temporal phase encoder. The encoded signal is filtered by multiple tunable notch filters (TNFs), which are composed of one or more cascaded FBG notch filters over mechanically stretching precision stages. The central wavelengths of the notch filters can be tuned by changing the tension of the FBGs. The filtered signal is decoded by a matched temporal phase decoder. To further investigate the impact of notch filtering on the TPC-OCDMA performance, an ASE light source is employed to change the OSNR of the system.

 

Fig. 1 Schematic of spectral notched TPC-OCDMA for optical steganography applications. TP: temporal phase; AC: autocorrelation; TNFs: tunable notch filters.

Download Full Size | PDF

Figure 2(a) and 2(b) show the optical spectra of the encoded signals with different notching wavelengths and the corresponding BER measurements (with or without additional ASE noise) respectively. As shown in Fig. 2(a), three notch filters with different central wavelengths (A: 1549.6nm, B: 1550.4nm, C: 1551.2nm) are employed in the experiments. The 3dB rejection bandwidth are 0.12nm, 0.14nm and 0.14nm, and the measured notch filter depth (with a spectral resolution of 0.06nm) is 13.2 dB, 14.9 dB, 14.5dB for notching filters A, B and C, respectively. Through mechanical stretching, a tuning range of ~3nm is achieved for the notch filtering for each FBG. Figure 2(b) compares the BER characteristics of the system with and without the added ASE noise.

 

Fig. 2 (Color online) (a) Spectra of notch filtered temporal phase coded signal without added ASE noise. The pink line shows the measured filtering curve of cascaded notching filters A, B and C, NF: notch filter. (b) measured BER results of TPC-OCDMA system with or without added ASE noise for different spectrum-notching cases (different central notching wavelengths and different number of notching wavelengths).

Download Full Size | PDF

Without extra ASE noise, the received optical power penalty compared with the case without notch filtering is within 0.12 dB for single notch filtering by A, B or C. The overlapped BER curves of the above three cases demonstrated that TPC-OCDMA signal is insensitive to the central notching wavelength.

When some ASE noise is loaded, notch filtering at the signal spectral edge can exclude more ASE noise component than the OCDMA signal. Better performance is observed when the spectral notching (with notch filter C: 1551.2nm) is at the signal spectral edge. Notching near the signal spectral center of the TPC-OCDMA signal (with notch filter A or B) causes larger power penalty than notching at the signal spectral edge (with notch filter C). The power penalty of system with three notch filters A, B and C [black line in Fig. 2(b)] is 0.35dB smaller than the one with two notch filters A and B [red line in Fig. 2(b)].

In a general coherent OCDMA system, multiple notch filterings could cause system performance degradation because of both pulse distortion and energy reduction [13]. Compared with the case without added ASE noise (high OSNR case), the notch filtering causes larger power penalties. With multiple notch filterings, the received optical power penalty for the TPC-OCDMA signals is less than 1.45dB, which shows that the TPC-OCDMA signal is robust against spectral notching filtering. Please note that here the notch filtering rejection bandwidth of 0.12nm-0.14nm is adequate to cover the spectral width of 10Gb/s NRZ-OOK signals.

3. Optical steganography transmission using spectral notched temporal phase coded optical signals and experiments

The experimental setup of the proposed approach of optical steganography based on spectral notched temporal phase coded optical signals is shown in Fig. 3 . The system includes one stealth channel and one public channel. The public channel is generated using external modulation of the continuous wave (CW) output from a DFB laser. The temporal phase coded optical signal is employed as the stealth channel. In order to suppress the interference of the public channel on the stealth channel, multiple tunable notch filters with 0.8nm spacing are employed before the temporal phase decoder.

 

Fig. 3 (Color online) Experimental setup of optical steganography transmission in WDM networks using spectral notched temporal phase coded signals. The inset eye diagram shows the measured eye diagram at point (a) when only the stealth signal is on. The inset spectra show the measured spectra at point (a) (in the case with and without stealth channel). The spectra without (red dot line) and with (black solid line) stealth signal are indistinguishable. MLFL: mode-locked fiber laser; PC: polarization controller; ATT: attenuator; TFF: thin film filter; EDFA: erbium doped fiber amplifier; MZM: Mach-Zehnder modulator.

Download Full Size | PDF

In the experiments, the optical pulse train generated by a mode-locked fiber laser (MLFL) has a pulse width of about 1.5ps. The repetition rate of the pulse train is 9.953GHz, and the central wavelength is tuned to 1550nm. The pulse train is gated down at four-to-one ratio and modulated to generate 2.488Gb/s (pseudo-random binary sequence at 2¹⁵-1 from data generator (1) optical signal. The signal is further amplified, encoded by the SSFBG and becomes noise-like signal in time domain. For the public channel, one DFB laser with central wavelength at 1549.63nm is modulated by a Mach-Zehnder modulator (MZM). The public channel is modulated at 2.488Gb/s (pseudo-random binary sequence at 2²³-1 from data generator (2). An ASE light source is loaded to emulate the ASE noise caused by the cascaded amplification in a typical WDM system. The combined signal from the three tributaries are amplified and received by a stealth user and a public user. As indicated in [1], the power ratio of the three tributaries should be optimized to minimize the impact of the stealth channel on the public channel while achieving error free transmission of the stealth signal. In our experiments, the measured power is: P_WDM = 6.8dBm, P_ASE = −4.4dBm, P_Stealth = −8.3dBm. For the stealth channel receiver, we use three TNFs (FBGs) to emulate notch filtering when there are multiple WDM channels. One TNF is tuned to suppress the interference from the public channel. The notched optical stealth signal is then decoded by a matched decoder. For the public channel receiver, the combined signal is filtered by a 0.6 nm thin film filter (as a demultiplexer in a WDM network) and then directly detected by a PD. The inset eye diagram shows the measured eye diagram at point (a) when only the noise-like stealth signal is on. The temporal profile of the encoded stealth signal is mainly determined by the code, the chip rate and the length of the SSFBG encoder [10–12]. With a longer SSFBG, the stealth signal will be spread over a longer period with lower power density which is preferable for optical steganography transmission. However, the maximum length of the SSFBG is limited by the data rate of the stealth channel due to the inter-symbol interference. The inset spectra in Fig. 3 show the spectra of the combined signals at point (a) (in the cases with and without the stealth channel) in the experimental setup. As shown in the inset spectra, the spectra of combined signal in the case without (red dot line) and with (black solid line) stealth signal are indistinguishable. The stealth signal is successfully concealed under ASE noise and public signal in the spectral domain.

Figure 4(a) –4(d) show the eye diagrams of the decoded signal without and with notch filtering. The stealth signal is degraded by the interference from public channel and ASE noise. The suppression of WDM interference by notch filtering is clearly shown when comparing Fig. 4(a) (without notch filtering) with Fig. 4(c) (with notch filtering). Figure 4(e), 4(f) show the measured eye diagrams of public channel for various cases with and without stealth signal and added ASE. Figure 4(e) and 4(f) show no obvious change in the eye diagrams of public channel for the overlaying of stealth channel, which shows that the stealth signal is concealed in the temporal domain.

 

Fig. 4 Measured eye diagrams: (a)-(b) received optical stealth signal without notch filtering; (c)-(d) received optical stealth signal with notch filtering. The interference of stealth channel from WDM channel is significantly reduced by notch filtering when comparing (a) with (c). (e)-(f) received public channel signal. The stealth signal is temporally concealed when comparing (e) with (f).

Download Full Size | PDF

Figure 5 shows the BER performance of the stealth channel and the public channel. Figure 5(a) shows the BER results of the stealth channel in four different cases. Error-free operation of the optical stealth channel can be achieved by notch filtering. The power penalty for the optical stealth channel (with a public WDM channel and with loaded ASE noise) is about 1.17 dB at BER = 1.0e-9, which shows that the spectral notching is effective in removing the WDM channel interference and causes relatively small degradation to the temporal phase coded optical stealth signal. Since we have employed multiple notch filtering and only one WDM channel in our experiments, additional power penalty is expected for stealth channel when more WDM channels exist. However, the limited bandwidth of stealth signal (10dB bandwidth: ~2.1nm) means limited number of spectrum-overlapped WDM channels (less than three 100GHz spacing WDM channels in our case). Therefore, the maximum power penalty caused by additional WDM channels will also be limited. Since the stealth channel can tolerate notch filtering at different wavelengths, the notch filters can be tuned to the dedicated wavelengths of the already exist WDM networks, e.g. ITUT grid, without large degradation of the stealth channel. Figure 5(b) shows the BER results of the public channel in these four cases. Because the temporal phase coded optical signal has a broad spectrum and a low spectral power density, with typical WDM channel filtering, e.g. 100GHz channel filtering, the extra power penalty of a public channel caused by the optical stealth signal is very small (~0.4dB in our experiments when ASE noise exists).

 

Fig. 5 Measured BER results of (a) received optical stealth channel. Multiple spectral notching is used when a public channel (a WDM channel) is present; (b) public channel. The power penalty induced by stealth signal is small (~0.4dB).

Download Full Size | PDF

4. Conclusion

A novel scheme for optical steganography in WDM networks using spectral notched temporal phase coded optical signals has been proposed and experimentally demonstrated. The approach takes the advantage of temporal phase coded optical stealth signal’s high tolerance to spectral notching. The experimentally demonstrated robustness of the present approach has implied its potential good compatibility with the existing WDM network for optical steganography applications. To our best knowledge, this is also the first report of spectral notched TPC-OCDMA application to optical steganography transmission.