Browsing by Author "Ee-Chien CHANG"

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Packet allocation for layered media transmission with path diversity
    (2003-11-01T00:00:00Z) Vu Thanh NGUYEN; Wei Tsang OOI; Ee-Chien CHANG
    Media streaming with path diversity can improve the quality of the stream received by reducing the variability of packet loss and delay in best-effort networks. In this paper, we show that the decision of which packets to send over which paths can greatly affect the quality of the streams received. We propose a general framework for analyzing transmissions of layered media stream with single sender, single receiver over two disjoint paths. Our analytical framework solves the following problem: Given a set of packets, dependencies between them and the aracteristics of the paths, how to distribute the packets over the paths (with possibility of duplications) to minimize the loss amount of useful data during transmission? Three different models to represent layered media are described and analyzed. We describe polynomial algorithms for finding optimal packet allocations in each model. Our results are significantly better than heuristic and packet allocation scheme used in previous work.
  • No Thumbnail Available
    Item
    Proactive Secret Sharing with Low Bandwidth and Fine Granularity
    (2003-11-01T00:00:00Z) Qiming LI; Charng Rang GUO; Ee-Chien CHANG
    We consider (K,N) proactive secret sharing schemes whereby a large number of secrets are shared by N servers, and the secrets are reconstructible from shares in any K servers. The schemes are proactive in the sense that the shares can be refreshed periodically to withstand mobile adversaries. Motivated by the difficulties encountered in practical handling of large number of secrets, we propose to look into two requirements. In contrast to the classical schemes, which require exchanges of linear number (w.r.t the number of secrets) of random bits during refreshment, we take the communication bandwidth into account and attempt to limit the number of random bits exchanged (low bandwidth requirement). Thus, each server has to refresh all shares based on the small number of random bits it receives. For practical considerations, each individual secret should be reconstructible without revealing other secrets (fine granularity requirement). The two requirements are conflicting because low bandwidth requirement imposes dependencies among shares, which might hinder fine granularity. Current known schemes achieve one or another, but not both. We propose a scheme that considers both requirements. The scheme also maintains many features found in classical schemes, for instance, it can be verifiable with minor modifications.