Silence Detection in Secure P2P VoIP Multiconferencing JOSE-VICENTE AGUIRRE 1 , RAFAEL ÁLVAREZ 2 , JULIA SÁNCHEZ 3 , ANTONIO ZAMORA 4 Dpt. of Computer Science and Artificial Intelligence University of Alicante Campus de San Vicente, Ap. Correos 99, 03080 Alicante SPAIN jaguirre@dccia.ua.es 1 , ralvarez@dccia.ua.es 2 , jsanchez@dccia.ua.es 3 , zamora@dccia.ua.es 4 This work was partially supported by the Spanish grants GV06/018 and MTM2005-05759 Abstract: - We analyze the impact of silence detection applied to secure P2P VoIP multiconferencing. Under the assumption that the number of simultaneous conversations is small regardless of the amount of participants we propose considering simple audio broadcasting as the basis of the multiconference scheme and study different alternatives regarding multiplexing, cryptographic primitives and implementation issues. Key-Words: - Multiconferencing, Broadcasting, VoIP, Audio, Security, Silence detection, P2P, Multiplex. 1 Introduction Most P2P audio multiconferencing systems focus on the worst case in order to achieve adequate performance and fair load sharing for all cases. The characteristics of human conversation make silence detection an effective way of eliminating background noise and reducing bandwidth requirements, since most of the time there will only be one active speaker even if the number of participants is very high. In this scenario, considering that audio multiconferencing will degenerate into a simple broadcast a significant amount of the time, is a suitable approach, extending this basic broadcasting to manage audio stream multiplexing. 2 Preliminaries We present here several key concepts which are necessary for the adequate understanding of the techniques introduced in the rest of the paper. 2.1 Mixing Mixing audio signals is the process of combining them into a new audio signal that sounds like all of them were played together. In the digital domain it is realized by adding each sample together, so for 2 signals , i i i M A B = + where i corresponds to the sample index. Conversely for n signals 1 2 . n i i i i M S S S = + + + " The main problem with digital mixing is that of the limited headroom: if we add too many loud signals we run the risk of exceeding the maximum values that can be represented with a certain bit depth, resulting in audio distortion. This problem can usually be minimized employing floating point arithmetic. 2.2 Gating A noise gate is a device or algorithm that shuts down an audio signal when the level is below a certain threshold. In this way, when a microphone is not being used, the noise gate can reduce the volume and virtually turn the microphone off, preventing any noise or undesired weak signal from being added to the mix. Noise gates have several parameters: − Threshold, or the level at which the gate begins to close or open. − Attack, or the time it takes for the gate to open. − Release, or the time it takes for the gate to close. The attack and release values can be constant or program dependent, varying in time as a function of the audio being processed. Proceedings of the 5th WSEAS Int. Conference on Information Security and Privacy, Venice, Italy, November 20-22, 2006 11