Loading...

Improvement of Security Features of Communication Networks with Network Coding

Mousavi, Nima | 2008

2538 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 39012 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Aref, Mohammad Reza; Salmasizadeh, Mahmoud
  7. Abstract:
  8. In this thesis, we study security features of the networks which use linear network coding to multicast information to some nodes of the network in presence of a wiretapper and also an active intruder. We consider information theoretic security in which we assume no computation limits for enemy. Thus, we take advantage of the essence of network coding instead of cryptographic tools to improve security features of the network. First, we show how providing shannon security for such networks in presence of wiretapper, leads to different cost in multicast rate, depending on initial network code designed. Based on this, we present a modified secure network code design which can provide perfect security against wiretapper. After that, we weaken the security condition and propose a new scheme to transmit information packets at multicast rate in such way that the probability of finding the content of a packet by wiretapper is very small. Then, we show that in presence of a wiretapper who has access to every links of the network with multicast rate of , one can divide symbols of the source and then send each of them in perfect secure manner, i.e, wiretapper has no information about each messages, even though she has information about all of them. We call this scheme as secure multiplex coding, and then propose a universal method based on MRD Linear codes, to implement it regardless of the network code used in the network. Next, we consider a network which is suffering from an active intruder who can inject erroneous packets to some links of the network. Although network coding is sensitive to such a pollution attack, we improve existing method to detect adversarial errors in the network with a higher probability. Then, we show that each adversarial node can only cause small fraction of potential adversarial error in the network and she also doesn’t know what error will be appeared in receivers. In addition, we present two achievable rates in which each receiver can detect erroneous packets with probability of one. These rates are dependent to network topology, and asymptotic on packets length
  9. Keywords:
  10. Network Coding ; Confidentiality ; Secure Network Coding ; Secure Multiplex Coding ; Wiretapper ; Active Enemy

 Digital Object List

 Bookmark