<?xml version="1.0" encoding="iso-8859-1"?>
<?xml-stylesheet href="paper.xsl" type="text/xsl"?>

<paper xmlns="http://www.cse.ucsd.edu/daniele/XML"
       xmlns:paper="http://www.cse.ucsd.edu/daniele/XML"
       xmlns:xhtml="http://www.w3.org/1999/xhtml">

  <filename>OptimalMulticast</filename>

  <title>Optimal communication complexity of generic multicast key
  distribution</title>

  <author>Daniele Micciancio</author>
  <author>Saurabh Panjwani</author>


  <reference>
    <link>http://www.ieee-acm-ton.org/</link>
    <journal>IEEE/ACM Transactions on Networking</journal>
    <year>2008</year>
    <volume>16</volume>
    <number>4</number>
    <pages>803-813</pages>
    <doi>10.1109/TNET.2007.905593</doi>
    <note>Preliminary version in Eurocrypt 2004</note>
  </reference>
  
  <abstract>
    <p xmlns="http://www.w3.org/1999/xhtml">
      We prove a tight lower bound on the communication complexity of secure 
      multicast key distribution protocols in which rekey messages are built 
      using symmetric-key encryption, pseudorandom generators and 
      secret sharing schemes.
      Our lower bound shows that the amortized cost of updating 
      the group key for each group membership change
      (as a function of the current group size)
      is at least <em>log<sub>2</sub>(n) - o(1)</em> basic rekey messages.
      This lower bound matches, up to a subconstant additive term,
      the upper bound due to Canetti, Garay, Itkis, 
      Micciancio, Naor and Pinkas <cite>[Proc. of Infocomm 1999]</cite>, 
      who showed that 
      <em>log<sub>2</sub>(n)</em> basic rekey messages (each time a user 
      joins and/or leaves the group) are sufficient. Our lower bound is, 
      thus, optimal up to a small, subconstant additive term.

      The result of this paper considerably strengthens previous lower bounds 
      by Canetti, Malkin and Nissim <cite>[Proc. of Eurocrypt 1999]</cite>
      and Snoeyink, Suri and Varghese 
      <cite>[Computer Networks 47(3):2005]</cite>,
      which allowed for neither the use of pseudorandom generators and 
      secret sharing schemes, nor the iterated (nested) application of the 
      encryption function. 
      Our model (which allows for arbitrarily nested combinations of encryption,
      pseudorandom generators and secret sharing schemes) is much more general, 
      and, in particular, encompasses essentially all known multicast key 
      distribution protocols of practical interest.
    </p>
  </abstract>

  <note>
    Preliminary versions in 
    <link doi="10.1007/b97182">Eurocrypt 2004</link>
  </note>
</paper>



