Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...
Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...
Page 2 Lecture Notes in Computer Science 2865 Edited by G. Goos ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
154 M. Just, E. Kranakis, and T. Wanthe network or malicious <strong>in</strong>siders from spoof<strong>in</strong>g or modify<strong>in</strong>g rout<strong>in</strong>g messages. Thedisadvantages are: 1) They are quite <strong>in</strong>efficient s<strong>in</strong>ce both the signature generation andverification process <strong>in</strong>volve the execution of computationally expensive functions. 2)They cannot prevent <strong>in</strong>ternal attacks.Given the <strong>in</strong>efficiency of digital signature mechanisms, some researchers [7, 27]proposed to use symmetric cryptographic primitives (i.e., one-way hash cha<strong>in</strong>s, onetimesignatures, authentication tree, etc.) for authenticat<strong>in</strong>g rout<strong>in</strong>g messages. Unfortunately,these approaches still do not prevent attacks from compromised <strong>in</strong>ternal routers.Hu, Johnson, and Perrig [13, 14] take the step further <strong>in</strong> secur<strong>in</strong>g distance vector rout<strong>in</strong>gprotocols <strong>by</strong> forc<strong>in</strong>g a node to <strong>in</strong>crease metrics when forward<strong>in</strong>g rout<strong>in</strong>g updatemessages. Therefore, their approaches can prevent compromised nodes from claim<strong>in</strong>gshorter distances. The disadvantage is that a malicious node can avoid traffic <strong>by</strong> claim<strong>in</strong>glonger distances.2.4 Prevent<strong>in</strong>g DoS Attacks on Data TrafficIt has been hypothesized that a network with QoS support can well resist DoS attackss<strong>in</strong>ce malicious packets will be dropped <strong>in</strong> the first place when fac<strong>in</strong>g network congestion.Other researchers proposed mechanisms [3, 6] to trace back to the orig<strong>in</strong> of themalicious packets which cause the network congestion and drop them <strong>in</strong> the routerswhere they first enter <strong>in</strong>to the victim network. Ingress/Egress filter<strong>in</strong>g can also be helpfulif IP spoof<strong>in</strong>g is utilized <strong>in</strong> the attack.Several approaches have been proposed to prevent DoS attacks on data forward<strong>in</strong>glevel. Perlman [22] proposed hop-<strong>by</strong>-hop packet acknowledgment to detect packetdropp<strong>in</strong>g <strong>in</strong> a network. The disadvantage is that it will generate significantly high rout<strong>in</strong>goverhead. Cheung et al [8] proposed a prob<strong>in</strong>g method for defeat<strong>in</strong>g denial of serviceattacks <strong>in</strong> a fixed rout<strong>in</strong>g <strong>in</strong>frastructure us<strong>in</strong>g neighborhood prob<strong>in</strong>g. It requires atest<strong>in</strong>g router to have a private address which allows it to generate a packet dest<strong>in</strong>ed toitself but goes through the tested router. This requirement is not practical <strong>in</strong> MANETs.A distributed monitor<strong>in</strong>g approach is proposed <strong>in</strong> [4] for detect<strong>in</strong>g disruptive routers.The protocol is based on the pr<strong>in</strong>ciple that any packets sent to a router and not dest<strong>in</strong>edto it are supposed to leave that router. This pr<strong>in</strong>cipal is not applicable to MANET dueto their chang<strong>in</strong>g network topology.Marti et al [19] proposed and implemented two protocols for detect<strong>in</strong>g and mitigat<strong>in</strong>gmisbehav<strong>in</strong>g nodes <strong>in</strong> wireless ad hoc networks <strong>by</strong> overhear<strong>in</strong>g neighborhoodtransmissions. Their method is very effective for detect<strong>in</strong>g misbehaviors <strong>in</strong> one-hopaway. To monitor the behavior of nodes two or more hops away, one node has to trustand rely on the <strong>in</strong>formation from other nodes, which <strong>in</strong>troduces the vulnerability thatgood nodes may be <strong>by</strong>passed <strong>by</strong> malicious or <strong>in</strong>correct accusation.Buchegger and Le Boudec [5] developed the CONFIDANT protocol for encourag<strong>in</strong>gnode cooperation <strong>in</strong> dynamic ad-hoc networks. Each node monitors the behaviorand ma<strong>in</strong>ta<strong>in</strong>s the reputation of its neighbors. The reputation <strong>in</strong>formation may be sharedamong friends. A trust management approach similar to Pretty GOOD Privacy (PGP)is used to validate received reputation <strong>in</strong>formation. Nodes with bad reputation may beisolated from the network. As a result, nodes are forced to be cooperative for their own