DoS attacks, whether they are intentional or unintended, are the most difficult VoIP-related threat to defend against. The packet switching nature of data networks allows multiple connections to share the same transport medium. Therefore, unlike telephones in circuitswitched networks, an IP terminal endpoint can receive and potentially participate in multiple calls at once. Thus, an endpoint can be used to amplify attacks. On VoIP networks, resources such as bandwidth must be allocated efficiently and fairly to accommodate the maximum number of callers. This property can be violated by attackers who aggressively and abusively obtain an unnecessarily large amount of resources. Alternatively, the attacker simply can flood the network with large number of packets so that resources are unavailable to all other callers.
In addition, viruses and worms create DoS conditions due to the network traffic generated by these agents as they replicate and seek out other hosts to infect. These agents are proven to wreak havoc with even relatively well-secured data networks. VoIP networks, by their nature, are exquisitely sensitive to these types of attacks. Remedies for DoS include logical network partitioning at layers 2 and 3, stateful firewalls with application inspection capabilities, policy enforcement to limit flooded packets, and out-of-band management. Out-of-band management is required so that in the event of a DoS event, system administrators are still able to monitor the network and respond to additional events.
Theft of services and information is also problematic on VoIP networks. These threats are almost always due to active attack. Many of these attacks can be thwarted by implementing additional security controls at layer 2. This includes layer 2 security features such as DHCP Snooping, Dynamic ARP Inspection, IP Source Guard, Port Security, and VLAN ACLs. The fundamental basis for this class of attacks is that the identity of one or more of the devices that participate is not legitimate.
Endpoints must be authenticated, and end users must be validated in order to ensure legitimacy Hijacking and call interception revolves around the concept of fooling and manipulating weak or nonexistent authentication measures. We are all familiar with different forms of authentication, from the password used to login to your computer to the key that unlocks the front door. The conceptual framework for authentication is made up of three factors: “something you have” (a key or token), “something you know” (a password or secret handshake), or “something you are” (fingerprint or iris pattern). Authentication mechanisms validate users by one or a combination of these. Any type of unauthenticated access, particularly to key infrastructure components such as the IP PBX or DNS server, for example, can result in disagreeable consequences for both users and administrators.
VoIP relies upon a number of ancillary services as part of the configuration process, as a means to locate users, manage servers and phones, and to ensure favorable transport, among others. DNS, DHCP, HTTP, HTTPS, SNMP, SSH, RSVP, and TFTP services all have been the subject of successful exploitation by attackers. Potential VoIP users may defer transitioning to IP Telephony if they believe it will reduce overall network security by creating new vulnerabilities that could be used to compromise non-VoIP systems and services within the same network. Effective mitigation of these threats to common data networks and services could be considered a security baseline upon which a successful VoIP deployment depends. Firewalls, network and system intrusion detection, authentication systems, anti-virus scanners, and other security controls, which should already be in place, are required to counter attacks that might debilitate any or all IP-based services (including VoIP services).
H.323 and SIP suffer security vulnerabilities based simply upon their encoding schemes, albeit for different reasons. Because SIP is an unstructured text-based protocol, it is impossibly to test all permutations of SIP messages during development for security vulnerabilities. Its fairly straightforward to construct a malformed SIP message or message sequence that results in a DoS for a particular SIP device. This may not be significant for a single UA endpoint, but if this “packet of death” can render all the carrier-class media gateway controllers in a network useless, then this becomes a significant problem. H.323 on the other hand is encoded according to ASN.1 PER encoding rules. The implementation of H.323 message parsers, rather than the encoding rules themselves, results in security vulnerabilities in the H.323 suite.
