SECURITY CONSIDERATIONS IN CURRENT VOIP
PROTOCOLS
Steffen Fries
Corporate Technology, Siemens AG, Otto-Hahn-Ring 6, 81730 Munich, Germany
Keywords: Voice over IP, security, encryption, authentication, integrity, SIP, H.323, Megaco, MGCP, SRTP.
Abstract: This document describes current state of the art security functionality provided in the four mainly used and
standardized Voice over IP (VoIP) signaling protocols, as there are the Session Initiation Protocol (SIP),
H.323, Megaco, and the Media Gateway Control Protocol (MGCP). It outlines the security provided by the
protocols itself or by dedicated security extensions including lower layer security protocols like Transport
Layer Security (TLS) or IPSec. Moreover, vulnerabilities, which still remain in protocols or certain scenar-
ios, are depicted as well. Furthermore discussed are also security approaches for the media data provided by
the Secure Real-time Transport Protocol (SRTP) and associated key management schemes. Conclusions are
given by identifying work areas, in which further security related work in the area of multimedia communi-
cation in general and VoIP in specific has to be done.
1 INTRODUCTION
Voice over IP is one of the driving factors for con-
vergent networks. It targets the migration from cur-
rent circuit switched networks to packet based net-
works for voice communication. VoIP is already
being used in enterprise and carrier environments for
VoIP trunking to connect legacy Private Branch
Exchange (PBX) via IP as well as for direct user
connectivity. Moreover VoIP is offered by public
service providers for residential customers.
Critical requirements on information security
cannot be satisfied by relying on “trust by wire secu-
rity” as done within traditional telephone network
architectures. The distributed and heterogeneous
VoIP system architecture itself enables attacks
against integrity and confidentiality of data commu-
nicated over the packet network. Inevitable threats to
the availability of the involved components are evi-
dent. Therefore comprehensive and consistent secu-
rity architectures are necessary prerequisites for
running VoIP communication systems based on
bundled multimedia standards. Note that in VoIP
scenarios signaling and media may traverse different
communication path and thus pose another challenge
to the overall security approach.
VoIP communication protocols and associated
security is currently being standardized mainly in
four standardization organizations. While the IETF
(Internet Engineering Task Force) and ITU-T (Inter-
national Telecommunication Union) define proto-
cols for VoIP like SIP, SRTP (IETF), and H.323
(ITU-T) as well as surrounding protocols, ETSI and
3GPP (Third generation Partnership Project) define
the architecture for NGN (Next Generation Net-
work) utilizing these protocols.
Typical threats to information security in general
and to VoIP in specific are eavesdropping or wire-
tapping, misuse of service (e.g., through masquerad-
ing), and manipulation.
An overview about information security mecha-
nisms in the commonly used signaling and media
protocols is provided as well as recently defined
security extensions to cover further use cases. It will
be shown that security mechanisms are already in-
corporated into the main VoIP protocols covering a
wide extend of known vulnerabilities.
Challenges for further security related work are
misuse of the multimedia protocol’s functionality for
SPIT (SPAM over Internet Telephony), Denial of
Service attacks (DoS). Moreover, besides basic call
more advanced features are going to be supported in
communication scenarios leading to further security
requirements to be met.
2 BASIC SCENARIOS AND
DEPLOYMENTS
VoIP is gaining more momentum and will be offered
in enterprises and public carriers for business and
183
Fries S. (2006).
SECURITY CONSIDERATIONS IN CURRENT VOIP PROTOCOLS.
In Proceedings of the International Conference on Security and Cryptography, pages 183-191
DOI: 10.5220/0002098601830191
Copyright
c
SciTePress
also personal use. The deployment of VoIP can be
done using different approaches providing also a
possible migration strategy.
A first step for introducing VoIP services is often
the trunking of voice connections between different
PBX’s via packet based networks as shown in
Figure 1. This option does not require the use of IP
enabled clients. Thus, VoIP can be introduced trans-
parent to the end user.
A next step in the deployment is consequently
the introduction of VoIP enabled networks in the
source and/or the target domain, including the ap-
propriate endpoints. These endpoints are able to
communicate completely over the packet based
network, while gateways ensure the connectivity to
legacy PBX systems.
IP Network
PSTN / PBX
Media
Gateway
H.323 / SIP
RTP
H.323 / SIP
RTP
Media
Gateway
PSTN / PBX
IP Network
PSTN / PBX
Media
Gateway
H.323 / SIP
RTP
H.323 / SIP
RTP
Media
Gateway
PSTN / PBX
Figure 1: VoIP Trunking.
A general scenario for VoIP is depicted in Figure
2 as an example for larger enterprises or for carrier
deployments. In these environments the media data
pose a high load on the gateways, when offering
connections to the Public Switched Telephone Net-
work (PSTN). Therefore multiple media gateways
can be coordinated by a single gateway controller.
The support for multiple gateways, targeting a
higher availability rate for connections to the PSTN
as well as cost reduction through the use of a single
controller for multiple gateways, will be discussed
with focus on the security of the supporting proto-
cols in section 5.3.
For smaller enterprises or home offices the me-
dia gateways and the media gateway controller col-
lapse to a single device, eliminating also the need for
additional gateway control protocols.
IP Network
PSTN
Media Gateway
Control Signaling
Megaco/H.248
or MGCP
Media
Gateway
RTP
SIP or H.323
RTP
SIP or H.323
RTP
SIP or H.323
RTP
SIP Proxy OR
H.323 Gatekeeper
AND Media
Gateway Controller
Client
Client
IP Network
PSTN
Media Gateway
Control Signaling
Megaco/H.248
or MGCP
Media
Gateway
RTP
SIP or H.323
RTP
SIP or H.323
RTP
SIP or H.323
RTP
SIP Proxy OR
H.323 Gatekeeper
AND Media
Gateway Controller
Client
Client
Figure 2: Multiple Gateway Support.
The final stage of VoIP would be a pure packet
based voice communication network and may be
seen as subset of the scenario depicted above. Be-
cause of this and as PSTN-based telephony will exist
in parallel to VoIP for quite some years, this sce-
nario is not further considered here.
3 TYPICAL VULNERABILITIES
AND THREATS
A threat to information security can be defined as
attempt at unauthorized access to an object by an
attacker.
Besides the general threats for packet-based IP
networks, within VoIP there are some dedicated
threat targets:
Identity: Attackers may spoof the identities of
service subscribers to misuse services and pro-
duce toll fraud. They may also spoof server iden-
tities to get access to user related information
(one form of spoofing server identities is com-
monly known as phishing).
Privacy: Eavesdropping or wiretapping of inse-
cure communication of signaling and/or media
connections may lead to loss of sensitive infor-
mation. Examples are PINs or passwords trans-
mitted over the signaling channel or (spoken)
personal information transmitted over the media
channel.
Availability: Voice or multimedia services are
susceptible to Denial of Service attacks, call hi-
jacking or call disruption. As multimedia-
services are real-time services, the importance of
this property increases. Also, if VoIP is used as a
SECRYPT 2006 - INTERNATIONAL CONFERENCE ON SECURITY AND CRYPTOGRAPHY
184
complete substitution for PSTN based telephony,
emergency call handling needs to be provided as
it will be required by regulation.
All of these properties are crucial for the general
acceptance of VoIP in business and personal com-
munication. Typical resulting security requirements
are depicted next.
4 SECURITY REQUIREMENTS
As VoIP is transmitted in (former) data networks,
similar security requirements as known for other
services in such a network apply. These are in the
first place:
Authentication: The property that the claimed
identity of an entity is correct.
Authorization: The process of giving someone
permission to do or have something.
Integrity: The property that information has not
been altered in an unauthorized manner.
Confidentiality: The property that information is
not made available or disclosed to unauthorized
individuals, entities or processes.
Additional challenges for VoIP security are
given through the different communication path for
signaling and media on one hand. Especially the
dynamic port assignment used for the media connec-
tions needs to cope with existing security infrastruc-
ture, like NAT devices and Firewalls.
On the other hand media security or, to be more
precise, the key management for the media security
is crucial (cf. also section 6.3). Here more complex
communication scenarios than simple point-to-point
connections need to be supported. In VoIP these
scenarios comprise the support for early media, i.e.,
a call initiator will receive media data prior to re-
ceiving an answer on the signaling path. Another
example is the forking of calls to reach different
endpoints simultaneously. These endpoints may
belong to the same user or form for instance a work-
ing group (commonly known is the ‘group pickup
feature’). Especially in the latter case user authenti-
cation and identity provision is required and may
pose potential obstacles.
5 VOIP SIGNALING
PROTOCOLS
Currently there exist several protocols for VoIP,
some of them are competitive, e.g., SIP and H.323,
and some are complementary, e.g., SIP and MGCP.
The following subsections comprise the discussion
of security focused on state of the art description of
the signaling protocols SIP (Rosenberg et al, 2002),
H.323 (ITU-T, 2003), Megaco/H.248 (Green, Ra-
malho and Rosen, 2000), and MGCP (Arango et al,
1999). Note, that the first two protocols provide
inherent security means, while the others mainly rely
on IPSec.
Common analyses show that the number of SIP
deployments increases more compared to H.323.
IPSec in general is applicable for UDP, TCP and
SCTP based signaling and may be used to provide
authentication, integrity and confidentiality for the
transmitted data. It supports end-to-end as well as
hop-by-hop scenarios. Thus, it may be used to pro-
vide security functions for all of the above protocols.
Nevertheless, it may not be applicable straight for-
ward, as it is often not considered by the signaling
standard itself and may not provide for advanced
communication scenarios. Moreover, due to the
message overhead of IPSec it may not be easily
applicable for real-time media communication.
Therefore alternative approaches have been taken.
5.1 SIP
The Session Initiation Protocol is specified by the
IETF in RFC3261 SIP (Rosenberg et al, 2002). It
enables the initiation and control of communication
sessions, which may be VoIP or complete multime-
dia sessions. The general architecture is shown in
Figure 2, were the server is called SIP proxy or redi-
rect server. The signaling of control information is
done using SIP, while media is sent using RTP.
SIP is a text–based Internet protocol similar to
protocols like Hypertext Transfer Protocol (HTTP)
and Simple Mail Transport Protocol (SMTP) in
contrast to other protocols like H.323 basing on the
Abstract Syntax Notation (ASN.1). ASN.1 requires
additional encoding and decoding operations. Never-
theless, if security using Secure Multipurpose Inter-
net Mail Extensions (S/MIME) is provided, ASN.1
becomes an issue also for SIP. SIP is independent of
the transport layer (supports User Datagram Protocol
(UDP), Transport Control Protocol (TCP) and
Stream Control Transmission Protocol (SCTP)) and
has been designed to be not restricted for Voice over
IP.
SIP already considers security and provides
measures for the most common scenarios. As SIP is
flexible with regard to extending the protocol, sev-
eral security enhancements have already been stan-
dardized. Furthermore, as new scenarios arise, new
SECURITY CONSIDERATIONS IN CURRENT VOIP PROTOCOLS
185
security extensions are being proposed. This is also
depicted in section 5.1.2.
Recent discussions also comprise the usage of
SIP to realize peer-to-peer VoIP systems. This will
pose new security requirements to VoIP. Main work
for security can be seen here in the area of distrib-
uted identities, prevention of spoofing, and denial of
service.
5.1.1 SIP Inherent Security Features
As stated above RFC3261 provides several security
features, which are depicted next.
Signalling data authentication using HTTP Digest
Authentication
SIP digest authentication is based on the HTTP
digest authentication defined in RFC2617 describing
a simple challenge-response paradigm. The remote
end is challenged using a nonce value. A valid re-
sponse contains a checksum (by default, the MD5
checksum) of the user name, the password, the given
nonce value, the HTTP method, and the requested
URI (Uniform Resource Identifier). In this way, the
password is never sent in the clear. Nevertheless,
there are some deficiencies in the usage of the HTTP
Digest scheme, as it does not provide complete mes-
sage integrity and may not be applied to all mes-
sages. Here TLS kicks in, which can be used to
enhance the signaling protection.
Approaches for protecting the signaling data
using TLS
RFC3261 mandates the support of TLS for SIP
server components (proxies, redirect servers, and
registrars) to protect SIP signaling. Using TLS for
User Agents (UAs) is recommended. TLS protects
SIP signaling messages against loss of integrity,
confidentiality and against replay. It provides inte-
grated key-management with mutual authentication.
TLS is applicable hop-by-hop between UAs and
proxies or between proxies. The SIP-Secure (SIPS)
scheme defined in RFC3261 requires the usage of
TLS to protect the signaling until the last proxy in
the call flow. Drafts are currently being discussed
addressing the deficiency regarding the last hop
usage of TLS (see section 5.1.2). TLS is usually
applied in a hop-to-hop fashion. Note that the calling
client does not get a confirmation about the usage of
TLS till the final recipient. It merely has to trust the
proxies acting according to the standard. The draw-
back of TLS in SIP scenarios is the requirement of a
reliable transport stack (TCP-based SIP signaling).
The recent development of Datagram-TLS (DTLS)
may provide for using a TLS-like protection also for
UDP based signaling. RFC3261 mandates the sup-
port of a dedicated TLS crypto scheme, which en-
sures interoperability. Other schemes may be negoti-
ated as part of the TLS handshake.
S/MIME to protect SIP message body data in an
end-to-end fashion
RFC3261 recommends the IETF defined stan-
dard S/MIME (RFC3850 and RFC3851) to be used
for end-to-end protection of signaling message pay-
loads. S/MIME within SIP supports the following
security services:
Authentication and Integrity Protection of Sig-
naling Data
Confidentiality of Signalling Data
Note that S/MIME is not widely used in current
SIP deployments and therefore not further discussed.
5.1.2 SIP Security Extensions
Meanwhile more complex SIP communication sce-
narios have been worked out, requiring additional
Request For Comments (RFCs) and new drafts en-
suring authentication and integrity and also confi-
dentiality. Identity is one of the intensively dis-
cussed topics within the SIP community. This can be
seen on the variety of documents concerning iden-
tity.
In the following an overview about a subset of
the most interesting security extensions is given.
These extensions solve different problems identified
in SIP communication scenarios.
SIP Authenticated Identity Body
SIP Authenticated Identity Body (AIB) defines a
generic SIP authentication token. It is provided by
adding an S/MIME body to a SIP request or re-
sponse in order to provide reference integrity over
its headers. The AIB is a digitally signed SIP mes-
sage or message fragment. This approach has been
standardized as RFC3893.
Enhancements for the Authenticated Identity
Management in SIP
The security options of RFC3261 utilize asym-
metric security in terms of certificates and corre-
sponding private keys. The distribution of these
credentials is often complex and limited to a dedi-
cated domain. Certificates can be used to provide
means for identity management, but this may not be
uniquely defined. An example can be drawn by two
different administrative domains with domain-
specific PKIs. The current draft (Peterson and
Jennings, WiP) addresses this limitation by defining
an authentication service providing assertions for the
user identity (Address of Record) transmitted in the
SECRYPT 2006 - INTERNATIONAL CONFERENCE ON SECURITY AND CRYPTOGRAPHY
186
header of SIP requests. This authentication service is
responsible for a dedicated domain.
Certificate Management Service for SIP
As stated before, certificate distribution is often
cumbersome and a global Public Key Infrastructure
(PKI) does not exist so far. Several security services
for SIP relay on certificates, so their distribution
becomes a crucial part. The draft (Jennings and
Peterson, WiP) describes a solution using a certifi-
cate server to provide certificates and even complete
user credentials using a Simple Authentication and
Security Layer (SASL) like approach. Users have
the option to store their complete credentials or only
the certificate over a secure connection on a central
server. This may be useful, when users are in need to
transmit credentials between different devices.
Connection Reuse / Outbound Connections
SIP defines the usage of TLS to protect the sig-
naling. But it requires only server components to
support mutual authentication. This leads to the
problem that clients without a TLS certificate cannot
receive inbound calls over TLS. This is due to the
fact that SIP communication is done over distinct
ports for inbound and outbound traffic. Furthermore,
an endpoint without a certificate TLS may not be run
in server-mode. To handle this deficiency two drafts
(Jennings and Hawrylyshen, WiP) and (Mahy, WiP)
describe the possibility of reusing already estab-
lished TLS connection, which were initiated by the
client. The basic idea is the provision of a flow iden-
tifier for the different streams within an established
session to identify inbound and outbound traffic.
5.2 H.323
H.323 (ITU-T, 2003) is an umbrella recommenda-
tion defined by the ITU-T (International Telecom-
munication Union). H.323 addresses call control,
multimedia management, and bandwidth manage-
ment as well as interfaces between LANs and other
networks. It includes point-to-point and multipoint
conferences. The first version has been defined in
1996 and has been improved consistently. Security
features have become part of the standard.
The general architecture is similar to SIP and can
also be depicted using Figure 2, were the server is
called H.323 gatekeeper. The signaling of control
information is done using H.225, while media is sent
using RTP.
The call establishment can be performed in vari-
ous ways, as there are gatekeeper routed calls, direct
routed calls with gatekeeper (for address resolution)
and plain direct routed calls.
The security functionality is defined within
H.235 (ITU-T, 2005). H.235 is splitted in 9 sub-
groups, describing so-called profiles for security in
H.323.
5.2.3 H.323 Security Profiles
The security profiles may be distinguished as pro-
files for signaling integrity and authentication and
for media security. They are related to the call mod-
els used in H.323. Note that the former H.235 an-
nexes have been reorganized in a new form
(H.235.x) recently.
H.235.0 provides the framework of the subse-
quent H.235 standards within H.323.
H.235.1 provides signaling integrity and authen-
tication using mutually shared secrets and keyed
hashes (HMAC-SHA1-96) in gatekeeper-routed
scenarios. This profile is widely implemented in
available H.323 solutions.
H.235.2 provides signaling integrity and authen-
tication using digital signatures on every mes-
sage in gatekeeper-routed scenarios. Since signa-
ture generation and verification is costly in terms
of performance, this profile may not gain mo-
mentum.
H.235.3 is a hybrid approach using both, H.235.1
and H.235.2. During the first handshake a shared
secret establishment is performed, protected by
digital signatures. Afterwards keyed hashes are
used for integrity protection, based on the estab-
lished shared secret.
H.235.4 is the adaptation of H.235.1 for direct
routed call scenarios with gatekeeper, were the
gatekeeper provides the key material for securing
the direct routed call.
H.235.5 specifies a framework for secure mutual
authentication during the registration and admis-
sion phase using weak shared secrets in combi-
nation with Diffie-Hellman key agreement for
stronger authentication during call signaling. Ex-
tensions to the framework to permit simultaneous
negotiation of TLS parameters for protection of a
subsequent call signaling channel are also pro-
vided.
H.235.6 describes the voice encryption profile.
This profile relies on certain security services as
part of the call signaling and connection setup
procedures; e.g., the Diffie-Hellman key agree-
ment and other key management functions and is
not compatible to SRTP.
H.235.7 can be seen as evolvement of H.235.6 as
it provides the framework for providing key ma-
SECURITY CONSIDERATIONS IN CURRENT VOIP PROTOCOLS
187
terial for media encryption based on MIKEY and
SRTP within H.235.
H.235.8 describes another approach for key
management for SRTP as ‘Key Exchange for
SRTP using secure signaling channels’. This is a
sdescriptions-like (cf. section 6.3.2) approach
were all parameter necessary for voice encryp-
tion are sent in-band protected by underlying se-
curity (e.g., like TLS) or by using Cryptographic
Message Syntax (CMS).
H.235.9 depicts the most recent extension of
H.235; the security gateway support for H.323.
The document defines a method for the discov-
ery of security gateways in the signaling path be-
tween communicating H.323 entities, and for
sharing of security information between a gate-
keeper and a security gateway in order to pre-
serve signaling integrity and privacy when cross-
ing network boundaries.
H.235 security features are defined to complete
H.323 scenarios. Additionally, the interworking with
security features of other multimedia protocol suites,
e.g., SIP, is considered. An example is the key man-
agement and media data encryption using H.235.7
and H.235.8.
5.3 Gateway Decomposition
Gateway decomposition describes the separation of
signaling and media functionality. The general goal
of gateway decomposition is a cost reduction
through the use of only one media gateway control-
ler responsible for several media gateways. The
media gateway controller connects to common mul-
timedia signaling protocols like SIP or H.323, de-
scribed above and controls the media gateways via a
trivial protocol. The general architecture approach is
shown in Figure 2.
There are two commonly used protocols within
the context of gateway decomposition – Megaco
(Green, Ramalho and Rosen, 2000) and MGCP
(Arango et al, 1999). Megaco/H.248 has basically
the same architecture as MGCP. Also the commands
are similar. However, the protocol models are quite
different.
Generally, these gateway control protocols pro-
vide complementary functionality and architectural
components to plain SIP or H.323 architectures.
5.3.1 Megaco
In June 1999 IETF Megaco Working Group (WG)
and ITU-T provided a unified document describing a
standard protocol for interfacing between Media
Gateway Controllers (MGCs) and Media Gateways
(MGs) Megaco/H.248. It is expected to win wide
industry acceptance as the official standard for de-
composed gateway architectures.
RFC2805 recommends security mechanisms that
may be in underlying transport mechanisms, such as
IPSec. H.248 goes even a step further by requiring
that IPSec shall be implemented, where the underly-
ing operating system and the transport network sup-
port IPsec. Implementations of the protocol using
IPv4 shall implement the interim AH scheme.
The interim AH scheme is the usage of an op-
tional AH header, which is defined in the H.248
protocol header. The header fields are exactly those
of the SPI, SEQUENCE NUMBER and DATA
fields as defined in RFC4302. The semantics of the
header fields are compliant with the "transport
mode" of RFC4302, except for the calculation of the
Integrity Check Value. The interim Authentication
Header (AH) scheme does not provide protection
against eavesdropping and replay attacks (the se-
quence number in the AH may overrun when using
manual key management since re-keying is not pos-
sible).
5.3.2 MGCP
MGCP is currently being maintained by Packet-
Cable
TM
and the Softswitch Consortium
TM
. In Octo-
ber 1999, MGCP was finally converted into an in-
formational RFC2705.
Regarding security, there are no mechanisms de-
signed into the MGCP protocol itself. The informa-
tional RFC2705 (Arango et al, 1999) refers to the
use of IPSec (either AH or Encapsulated Security
Payload (ESP)) to protect MGCP messages. Without
this protection an adversary could setup unauthor-
ized calls or interfere with ongoing authorized calls.
Beside the usage of IPSec, MGCP allows the call
agent to provide gateways with session keys that can
be used to encrypt the payload of the Real-time
Transport Protocol (RTP), protecting against eaves-
dropping. To achieve this RTP encryption, described
in RFC3550 (Schulzrinne, 2003) may be applied.
Session keys can be transferred between the call
agent and the gateway by using SDP Handley and
Jacobson, 1998) either directly or using dedicated
key management extensions.
6 MEDIA DATA SECURITY
The signalling protocols described above do not
consider the encryption of media data directly. Nev-
ertheless, they allow the negotiation of security
SECRYPT 2006 - INTERNATIONAL CONFERENCE ON SECURITY AND CRYPTOGRAPHY
188
features and also the key management for the secu-
rity associations of the media channels.
Real-time media data in VoIP environments is
usually transmitted using RTP. For RTP basically
two approaches for security provisioning exist,
which are discussed in the following.
6.1 RTP/RTCP
RFC3550 defines the RTP protocol as well as the
associated Real Time Control Protocol (RTCP).
Using the optional RTP encryption as defined in
RFC3550 provides for confidentiality for media data
using the Data Encryption Standard (DES) as default
method. The functionality is limited as the following
issues are not considered in the RTP standard:
key management
authentication and message integrity (not feasi-
ble without a key management infrastructure)
replay protection
Because of these points, the usage of this option
is not widely deployed.
6.2 SRTP/SRTCP
The RTP standard provides the flexibility to adapt to
application specific requirements with the option to
define profiles in companion documents. SRTP
(Baugher et al, 2004) has been recently defined as
RTP profile and addresses the limitations. This pro-
file provides confidentiality using the Advanced
Encryption Standard (AES, the successor of the
DES), message authentication (using keyed hashes)
and replay protection to the RTP/RTCP traffic.
SRTP does not define an own key management
and relies on solutions like MIKEY or sdescriptions
(cf. section 6.3). The key management messages
have to be transported by the signaling protocol.
For SDP based protocols like SIP or MGCP the
key transport can be transported as part of SDP it-
self, while for H.323 extensions to the base protocol
are defined through H.235.7 and H.235.8 as de-
scribed above.
6.3 Key Management
In contrast to common data applications, the key
management for negotiating the VoIP media security
has to cope with several additional requirements, as
there the real-time conditions, support of advanced
communication scenarios like forking, retargeting
and also conferencing. Moreover, due to the nature
of VoIP to separate signaling and media traffic, the
key management may be done along the path of the
signaling data, the media data, or combined.
Currently there is a broad discussion about the
key management for SRTP using SIP within the
IETF as meanwhile 13 different approaches exist.
They utilize different cryptographic techniques like
pre-shared keys, Diffie-Hellman key agreement,
digital signatures, or asymmetric encryption (cf.
(Audet and Wing, WiP)). Due to the different
mechanisms used, the approaches are not interoper-
able. Moreover, none of the current approaches
provides a solution for all of the scenarios stated
above.
In the following subsection two promising key
management approaches out of the 13 are depicted
in more detail. Both belong to the cluster of signal-
ing path key management approaches and have been
deployed already. Examples for media path key
management are given through ZRTP (cf.
(Zimmermann, WiP)) and DTLS usage (cf.
(McGrew and Rescorla, WiP)).
6.3.1 Multimedia Internet Keying
MIKEY (Multimedia Internet Keying) has been
defined in the IETF within RFC3830 (Akko, 2004).
It defines an authentication and key management
framework that can be used for real-time applica-
tions (both for peer-to-peer communication and
group communication). In particular, RFC3830 is
defined in a way to support SRTP in the first place
but is open to enhancements to be used for other
purposes too. MIKEY has been designed to meet the
requirements of initiation of secure multimedia ses-
sions. Such requirements are for instance the estab-
lishment of the security parameters for the multime-
dia protocol within one round trip.
Another requirement is the provision of end-to-
end keying material, and also independence from
any specific security functionality of the underlying
transport layers.
MIKEY defines several options for the user au-
thentication and negotiation of the master keys with
a maximum of a complete round trip as there are:
Symmetric key distribution (based on pre-shared
keys and keyed hashes), which may proceed with
a single message.
Asymmetric key distribution (based on asymmet-
ric encryption) may proceed with a single mes-
sage.
Diffie Hellman key agreement protected by digi-
tal signatures (complete roundtrip)
Unprotected key distribution, i.e., without au-
thentication, integrity, or encryption, is also possi-
SECURITY CONSIDERATIONS IN CURRENT VOIP PROTOCOLS
189
ble, but not recommended without any underlying
security like TLS or similar. This use case is compa-
rable with the sdescription approach described be-
low.
Two MIKEY enhancements exist as drafts,
which are likely to advance to a RFC soon.
Diffie Hellman key agreement protected by
symmetric pre-shared keys and keyed hashes
Asymmetric (encrypted) key distribution with in-
band certificate provision
The default and mandatory key transport encryp-
tion is the AES in counter mode, where MIKEY
references RFC3711. MIKEY uses a 160-bit authen-
tication tag, generated by HMAC with SHA-1 as
mandatory algorithm described in the associated
RFC2104. Also mandatory, when asymmetric
mechanisms are used, is the support of X.509v3
certificates for public key encryption and digital
signatures.
Recently the usage of elliptic curves has been
proposed targeting performance saving and enabling
the use of shorter cryptographic key material by
keeping the same level of security compared to the
currently used RSA.
6.3.2 Security Descriptions
Besides MIKEY a second key management for
SRTP has been proposed in the IETF, which utilizes
the plain Session Description Protocol (SDP). The
approach is based on the offer answer model of SDP
and transmits all necessary SRTP parameter in a
new attribute field and is called sdescriptions (cf.
(Flemming and Baugher, WiP)).
The protection of this field is left to SIP itself
(applying S/MIME in an end-to-end fashion) or may
be done using TLS (in a hop-to-hop fashion) or even
IPSec. It therefore nicely integrates with SIP. For
other signaling protocols like H.323 there exist simi-
lar approaches (through H.235.8). Because of the
signaling protocol dependent approach, this solution
lacks the support of end-to-end security.
7 CONCLUSION
As shown in this paper, current multimedia proto-
cols already consider security to certain extends.
SIP and H.323 are here the most advanced proto-
cols, as they provide inherent security measures for
user authentication and message integrity. Confiden-
tiality can be achieved by additional measures like
S/MIME or underlying security protocols TLS or
IPSec. Both signaling protocols also provide options
to transport a key management for SRTP. Several
key management approaches have been proposed
leading to the requirement for profiles to ensure easy
(inter)operation.
SIP and H.323 suffer from dynamic port signal-
ing as part of the protocol payload, which may lead
to problems with widely deployed NAT devices,
when message integrity or confidentiality is desired.
Within the IETF efforts have been spent to cope
with this problem by providing a methodology for
Interactive Connectivity Establishment (ICE,
(Rosenberg, WiP)).
Security for gateway control protocols is mainly
provided by using IPSec as underlying security
protocol. Here the communication association is
merely between the controller and the associated
gateways. Thus, the signaling scenarios are rather
simple compared to SIP and H.323.
As the scenarios, in which VoIP is about to be
deployed, are getting more complex through the
integration of already available features of the leg-
acy telephone system, new security requirements
arise. An example is the handling of security proper-
ties like user authentication or media session confi-
dentiality in case of call transfer. This is especially
becoming interesting in scenarios were the partici-
pants of a call cannot rely on the same security in-
frastructure. Here a global PKI solution could sup-
port user authentication. As this is not available right
now, alternative approaches are necessary.
Potential obstacles for VoIP usage may arise
through the possibility of misusing the infrastructure
resulting in Denial of Service. SPIT will pose an-
other potential obstacle. As seen for today’s email
communication SPAM poses a severe problem for
communication. To counter similar threats in VoIP
certain measures have to be taken and are already
being discussed.
Further challenges are given through the user’s
request for interoperability, whereby this relates to
several facets, like the implementation of a standard
through different vendors and also proprietary en-
hancements. Moreover other solutions are available,
which are not standardized so far. While the first
point may be covered by regular interoperability
tests, the second leads to a subset of common func-
tionality, which ensures interworking. For SIP the
definition of this common set of functions is done in
the SIP Forum. A prominent example for the third
point is Skype, providing security of signaling and
media traffic in a proprietary way, not interoperable
to SIP or H.323 based clients. Thus security inter-
working in an end-to-end fashion may not be possi-
ble.
SECRYPT 2006 - INTERNATIONAL CONFERENCE ON SECURITY AND CRYPTOGRAPHY
190
Last but not least, security functions provided by
signaling and media protocols are only one part of
the multimedia puzzle. There is more work to be
done in defining the suitable system architecture and
deploying multimedia systems in a secure way. Most
important, multimedia systems have to be designed
into already existing networks under consideration
of security, availability, quality of service, and also
legal terms. This is also outlined in (Kuhn, Walsh
and Fries, 2005). Especially for emergency services
appropriate measures have to be taken to meet the
balance between security and availability.
ACKNOWLEDGEMENTS
The author would like to thank Wolfgang Klasen
and Michael Montag for their valuable review com-
ments and discussions.
REFERENCES
Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston,
A., Peterson, J., Sparks, R., Handley, M. and Schooler,
E., 2002, RFC3261: SIP: Session Initiation Protocol
ITU-T, 2003, H.323v5: Packet-based multimedia commu-
nications systems
Rosenberg, J., Work in Progress, ICE: A Methodology for
Network Address Translator (NAT) Traversal for Of-
fer/Answer Protocols,
Greene, N., Ramalho, M. and Rosen, B., 2000, RFC2805:
Media Gateway Control Protocol Architecture and
Requirement,
Arango, M., Dugan, A., Elliott, I., Huitema, C. and
Pickett, S., 1999, RFC2705: Media Gateway Control
Protocol Version 1.0
Handley, M. and Jacobson, V., 1998, RFC2327: SDP:
Session Description Protocol
Schulzrinne, H., Casner, S., Frederick, R. and Jacobson,
V., 2003, RFC3550: RTP: A Transport Protocol for
Real-Time Applications
Baugher, M., McGrew, D., Naslund, M., Carrara, E. and
Norrman, K., 2004, RFC3711: The Secure Real-time
Transport Protocol
Audet, F. and Wing, D., Work in Progress, Evaluation of
SRTP Keying with SIP
Arkko, J., Carrara, E., Lindholm, F., Naslund, M. and
Norrman, K., 2004, RFC3830: MIKEY: Multimedia
Internet KEYing
ITU-T, 2005, H.235.0: Security framework for H-series
Arkko, J., Carrara, E., Lindholm, F., Naslund, M. and
Norrman, K., Work in Progress, Key Management Ex-
tensions for Session Description Protocol (SDP) and
Real Time Streaming Protocol (RTSP)
Peterson, J. and Jennings, C., Work in Progress, En-
hancements for Authenticated Identity Management in
the Session Initiation Protocol
Jennings, C. and Peterson, J., Work in Progress, Certifi-
cate Management Service for The Session Initiation
Protocol,
Jennings, C. and Hawrylyshen, A., Work in Progress, SIP
Conventions for UAs with Outbound Only Connection,
Mahy, R., Work in Progress, Connection Reuse in the
Session Initiation Protocol (SIP)
Flemming, A., Baugher, M. and Wing, D., Work in Pro-
gress, Session Description Protocol Security Descrip-
tions for Media Streams
Zimmermann, P., Work in Progress, ZRTP: Extensions to
RTP for Diffie-Hellman Key Agreement for SRTP
McGrew, D. and Rescorla, E., Work in Progress, Data-
gram Transport Layer Security Extension to Establish
Keys for Secure Real-time Transport Protocol
Kuhn, D.R., Walsh, T.J. and Fries, S., 2005, Security
Considerations for Voice over IP Systems, SP800-58,
US NIST
SECURITY CONSIDERATIONS IN CURRENT VOIP PROTOCOLS
191