a1495_1.pdf JWA83.pdf

TDM-PON Security Issues: Upstream Encryption is Needed David Gutierrez, Jinwoo Cho and Leonid G. Kazovsky Photonics and Networking Research Laboratory, Stanford University, 058 Packard Building, Stanford, California 94305, USA [email protected]

Abstract: TDM-PONs (E/B/GPON) present several security issues that can easily be exploited by malicious users. We summarize these issues and present experimental results to demonstrate that, in particular, upstream encryption is required to prevent eavesdropping. 2007 Optical Society of America OCIS codes: (060.2330) Fiber optics communications, (060.4250) Networks

1. Introduction Time Division Multiplexed Passive Optical Networks (TDM-PONs), in its three versions, Ethernet, Broadband and Gigabit PON (EPON, BPON and GPON) are currently being deployed in Asia, North America and, to a lesser degree, Europe. Hundreds of thousands of home users already enjoy the increased bandwidth that TDM-PONs provide with respect to DSL or Cable, and the expected number of users is expected to be in the tens of millions by 2010 [1]. Every day, homes and small businesses will rely more and more on these networks for financial transactions, private communications and even telemedicine. This creates a strong requirement for access networks to be trustworthy, secure and reliable. The IEEE 802.3ah EPON standards do not specify any authentication and encryption mechanisms. Thus, particular proprietary solutions have been implemented by EPON manufacturers [2]. The ITU G.983 BPON recommendations do not specify particular security mechanisms either, but its successor, the ITU G.984 GPON recommendations do use the Advanced Encryption Standard (AES) for downstream transmission. We believe there are several security vulnerabilities in TDM-PONs. In the following sections, we briefly explain these issues, provide an experiment in which we illustrate one of these problems and conclude that, in particular, upstream encryption is required to prevent eavesdropping. 2. Security Issues in TDM-PON The main security problems in TDM-PON are: (1) denial of service attacks, (2) eavesdropping and (3) masquerading of an ONU. (1) A simple denial of service attack can take place if a malfunctioning or purposefully corrupted upstream laser diode at an Optical Network Unit (ONU) is set to continuously transmit at the upstream wavelength with a high enough power to block all other ONUs from getting their own data through. Since the network is passive, it is quite hard to discover the problematic ONU and disable it or its connection port. Some mechanisms have been proposed to effectively identify the attacker and/or disconnect it from the network [3, 4]. (2) Eavesdropping happens when an ONU is able to listen to the data that is sent to or from another ONU. Some of the EPON authentication and encryption mechanisms proposed in the literature [5], as well as the GPON standards assume that the upstream traffic of a particular ONU cannot be observed by the other ONUs due to the high directionality of the components of the Optical Distribution Network (ODN). This means that even though all ONUs can receive all the downstream traffic which has been encoded by the Optical Line Terminal (OLT), in theory no ONU should be able to detect the upstream traffic of neighboring ONUs. The security model assumes that upstream transmission is secure from eavesdropping and therefore doesn’t require encryption. Furthermore, that the keys to encrypt the downstream data can be sent in the clear upstream since this is a secure medium. These keys are subsequently used by the OLT to encrypt downstream data differently for ech ONU. The ODN directionality assumption is questionable in practice. As has been reported by some network operators in the U.S., the measured reflections for dirty connectors and splitters are the following [6]:

ORL

Table 1. Optical Return Loss for UPC and APC connectors and splitters [6]. Open / Dirty Dirty UPC Splitter APC Splitter UPC Connector APC Connector ~ 15 dB ~ 22 – 31 dB ~ 33 dB ~ 55 dB

a1495_1.pdf JWA83.pdf

From the experience of one of the authors of this paper, it is not uncommon to find dirty connectors affecting the reach and quality of deployed PONs. Given the ORLs mentioned in the Table 1, it would be therefore not uncommon either to have situations where the reflections are enough for a Malicious ONU to observe the upstream traffic of a Victim ONU, but not enough to prevent upstream transmission from happening, making the problem unnoticeable to the network operator. Furthermore, the Malicious ONU could even obtain the keys to decrypt the downstream transmission of the Victim ONU as well. (3) Masquerading happens when a Malicious ONU poses as the Victim ONU. Following the argument above, if the Malicious ONU has access to all the upstream and downstream information of the Victim ONU, it can effectively identify itself as the Victim ONU. 3. Experimental Setup and Results In the experimental setup shown in Fig. 1 we illustrate how a Malicious ONU can easily eavesdrop on the upstream traffic of a Victim ONU thanks to the ODN reflections. At the Victim ONU we have a laser transmitter at 0 dBm in the upstream direction with data modulated by a PPG at 155 Mbit/s with a PRBS 223-1 bits long. The upstream rate of 155 Mbit/s was chosen to match that of current TDM-PON deployments, as of 2006. This data travels through 1.6 Km of SMF until the nearest passive coupler/splitter of the ODN. Also attached to it by a 1.2 km SMF is the Malicious ONU, which uses any reflections from the ODN to read the upstream data from the Victim ONU. We use a receiver with a sensitivity of approximately -37.2 dBm for a BER of 10-9 as is shown in the Back-to-Back BER measurements on Fig. 2. For the Passive Coupler, we use two different setups. In Setup 1, we use three two-window 1x2 50/50 fused FBT couplers terminated with UPC connectors and with a directivity ≥ 55 dB according to the specs. As mentioned before, depending on the type of termination that we use for this coupler in the upstream direction, the amount of reflected power received by the Malicious ONU will differ. In Setup 1, we leave the upstream connector of the Passive Coupler unterminated with a UPC Connector. In this case, the reflected power is about 28.8 dBm and the Malicious ONU can detect all the upstream information of the Victim ONU virtually error-free.

1:4 Passive Coupler

LD [TXOP]

Setup 1 1.6 Km SMF

Driver

2x1 Couplers

Victim ONU Unterminated UPC upstream connector PD

Setup 2

Limiting Amp.

[RXEL]

PPG BERT

2x2 Couplers

1.2 Km SMF 50Ω Ω

APC Connector terminated with unidirectional attenuator

155 Mbit/s PRBS 223-1

Malicious ONU Figure 1. Experimental setup.

Due to the high reflectivity of UPC connectors, some PON deployments use APC connectors. Setup 2 reflects this situation. We use three two-window 2x2 50/50 fused FBT couplers terminated with angled connectors and with a directivity ≥ 65 dB according to the specs. Furthermore, we terminate the upstream connector of the passive coupler with a unidirectional 20 dB attenuator to emulate an upstream connection. Even in this situation, the -33.0 dBm reflected power is again enough for the Malicious ONU to eavesdrop on the Victim ONU. Fig. 2 illustrates the BER measurements for this experiment. The Malicious ONU receiver has a sensitivity of approximately -37.2 dBm for a 10-9 BER according to our back-to-back measurements. In both Setup 1 and Setup 2, the power the Malicious ONU receives is much higher than the one needed even for a 10-11 BER. Please note

a1495_1.pdf JWA83.pdf

that even though this reflected power may be affected by the quality and cleanliness of our connections, the measured ORL is even less than the one reported in [6] for dirty connectors in the field. The two eye diagrams correspond to the optical transmission and electrical receiver for Setup 2 as shown in Fig. 1 ([TXOP] and [RXEL]). Table 2. Reflected power from Victim ONU to Malicious ONU Victim ONU Pout (dBm) 1:4 Coupler Termination Malicious ONU Pin (dBm) 0 UPC - 28.8 0 APC + Unidirectional Attenuator - 33.0

Back-to-Back Setup 1 Setup 2 1 ns

[TXOP]

1 ns

[RXEL]

Fig. 2. BER measurements en eye diagrams.

To obtain the BER curves we used an attenuator between the fiber and the photodiode of the Malicious ONU. Note that in both cases, the actual received power in Table 2, without the attenuator for BER measurements, was much higher than what is needed for 10-11 BER. Newer PON deployments will use higher bit rates than the one we used for this experiment. In this case, however, it would be possible to use an APD receiver with a high-precision electrical amplifier to detect the data from the reflections just as we did in this experiment. 4. Conclusion TDM-PONs (E/B/GPON) present several security issues that can easily be exploited by malicious users, including Denial of Service attacks, eavesdropping and ONU masquerading. We present an experiment in which a Malicious ONU can accurately eavesdrop on the upstream traffic of a Victim ONU that shares a passive coupler at the ODN. Given that the keys to encode downstream traffic are transmitted in the clear in the upstream direction, the Malicious ONU could also potentially decode the downstream traffic of the Victim ONU. In conclusion, the proposed authentication and encryption mechanism should not assume that upstream TDM-PON transmission is secure. 5. References [1] [2] [3] [4] [5] [6]

R. Rubenstein, A. Armstrong, B. Baker, “PON IC Opportunities Expand as Market Ramps”, RHK Market Forecast: Annual, June 2004. G. Kramer, “What is Next for Ethernet PON?”, Proceedings of COIN 2006, July 2006. Y. Horiuchi, N. Edagawa, “ONU Authentication Technique Using Loopback Modulation within a PON Disturbance Environment”, Proceedings of OFC 2005, OFI3. S. Wong, W.-T. Shaw, S. Das, L.G. Kazovsky, “Enabling Security Countermeasure and Service Restoration in Passive Optical Networks”, Proceedings of IEEE Globecom 2006. Y. Meng, T. Jiang, D. Xiao, “Analysis and Solutions of Security Issue in Ethernet PON”, Proceedings of SPIE, Vol. 5626, 2005. Vincent O’Byrne, “Verizon’s Fiber to the Premises: Lessons Learned”, Proceedings of OFC 2005, OWP6.