As Voice over IP becomes more prevalent, value-adds to the service will become ubiquitous. Voice over IP (VoIP) is no longer a single service application, but an array of marketable services of increasing depth, which are moving into the non-desktop market. In addition, as the range of devices being generally used increases, it will become necessary for all services, including VoIP services, to be accessible from multiple platforms and through varied interfaces. With the recent introduction and growth of the open source software PBX system named Asterisk, the possibility of achieving these goals has become more concrete. In addition to Asterisk, a number of open source systems are being developed which facilitate the development of systems that interoperate over a wide variety of platforms and through multiple interfaces. This thesis investigates Asterisk in terms of its viability to provide the depth of services that will be required in a VoIP environment, as well as a number of other open source systems in terms of what they can offer such a system. In addition, it investigates whether these services can be made available on different devices. Using various systems built as a proof-of-concept, this thesis shows that Asterisk, in conjunction with various other open source projects, such as the Twisted framework provides a concrete tool which can be used to realise flexible and protocol independent telephony solutions for a small to medium enterprise.

When adding Confidentiality, Integrity and Availability (CIA) to a multi-user VoIP (Voice over IP) system, performance and quality are at risk. The aim of this study is twofold. Firstly, it describes current methods suitable to secure voice streams within a VoIP system and make them available in an Asterisk-based VoIP environment. (Asterisk is a well established, open-source, TDM/VoIP PBX.) Secondly, this study evaluates the performance cost incurred after implementing each security method within the Asterisk-based system, using a special testbed suite, named DRAPA, which was developed expressly for this study. The three security methods implemented and studied were IPSec (Internet Protocol Security), SRTP (Secure Real-time Transport Protocol), and SIAX2 (Secure Inter-Asterisk eXchange 2 protocol). From the experiments, it was found that bandwidth and CPU usage were significantly affected by the addition of CIA. In ranking the three security methods in terms of these two resources, it was found that SRTP incurs the least bandwidth overhead, followed by SIAX2 and then IPSec. Where CPU utilisation is concerned, it was found that SIAX2 incurs the least overhead, followed by IPSec, and then SRTP.

Voice-enabled applications, applications that interact with a user via an audio channel, are used extensively today. Their use is growing as speech related technologies improve, as speech is one of the most natural methods of interaction. They can provide customer support as IVRs, can be used as an assistive technology, or can become an aural interface to the Internet. Given that the telephone is used extensively throughout the globe, the number of potential users of voice-enabled applications is very high. VoiceXML is a popular, open, high-level, standard means of creating voice-enabled applications which was designed to bring the benefits of web based development to services. While VoiceXML is an ideal language for creating these applications, VoiceXML gateways, the hardware and software responsible for interpreting VoiceXML applications and interfacing with the PSTN, are still expensive and so there is a need for a low-cost gateway. Asterisk, and open-source, TDM/VoIP telephony platform, can be used as a low-cost PSTN interface. This thesis investigates adding a VoiceXML service to Asterisk, creating a low-cost VoiceXML prototype gateway which is able to render voice-enabled applications. Following the Component-Based Software Engineering (CBSE) paradigm, the VoiceXML gateway is divided into a set of components which are sourced from the open-source community, and integrated to create the gateway. The browser requires a VoiceXML interpreter (OpenVXI), a Text-To-Speech engine (Festival) and a speech recognition engine (Sphinx 4). The integration of the components results in a low-cost, open-source VoiceXML gateway. System tests show that the integration of the components was successful, and that the system can handle concurrent calls. A fully compliant version of the gateway can be used in the real world to render voice-enabled applications at a low cost. / KMBT_363 / Adobe Acrobat 9.55 Paper Capture Plug-in