13.07.2015 Views

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 ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Resist<strong>in</strong>g Malicious Packet Dropp<strong>in</strong>g <strong>in</strong> Wireless Ad Hoc Networks 155<strong>in</strong>terest. Our proposed prob<strong>in</strong>g technique can be used as one of the monitor<strong>in</strong>g techniques<strong>in</strong> the CONFIDANT protocol.Awerbuch et al [2] proposed a secure rout<strong>in</strong>g protocol for resist<strong>in</strong>g <strong>by</strong>zant<strong>in</strong>e failures<strong>in</strong> a wireless ad hoc network. The protocol requires an ultimate dest<strong>in</strong>ation to sendan acknowledgment back to the sender for each of its successfully received packets. Ifthe loss rate of acknowledgment packets exceeds the predef<strong>in</strong>ed threshold, which is setto be slightly above the normal packet loss rate, the route used for send<strong>in</strong>g packets fromthe source to the dest<strong>in</strong>ation is detected as faulty and a b<strong>in</strong>ary search prob<strong>in</strong>g techniqueis deployed to locate the faulty l<strong>in</strong>k. The disadvantages of this protocol are: 1) it may<strong>in</strong>cur significant rout<strong>in</strong>g overhead; 2) a data packet with an <strong>in</strong>serted probe list can bedist<strong>in</strong>guished from those without probe lists, although the probe list is onion encryptedand cannot be tampered en route. Our proposed prob<strong>in</strong>g technique differs <strong>in</strong> that it canbe implemented above the network layer (e.g., based on UDP), and the end-to-end encryptionof IP payload us<strong>in</strong>g pair-wise shared keys can prevent <strong>in</strong>termediate nodes fromdist<strong>in</strong>guish<strong>in</strong>g probe messages from data packets.Padmanabhan and Simon [20] proposed a secure traceroute to locate faulty routers<strong>in</strong> wired networks. In their approach, end hosts will monitor network performance. If anend-to-end performance degrade is detected <strong>by</strong> a host to a dest<strong>in</strong>ation, a compla<strong>in</strong>t bitis set <strong>in</strong> all the subsequent traffic to that dest<strong>in</strong>ation. The compla<strong>in</strong><strong>in</strong>g host itself or therouter sitt<strong>in</strong>g closest to the compla<strong>in</strong><strong>in</strong>g host may start the troubleshoot<strong>in</strong>g process if itobserves enough compla<strong>in</strong>ts. It first sends a secure traceroute packet to the next hop,which can be derived from its rout<strong>in</strong>g table. The router receiv<strong>in</strong>g the secure traceroutepacket is expected to send a response back which also <strong>in</strong>cludes a next hop address. Thisprocess repeats until a faulty router is located (no response received from it) or everyrouter on the path to the ultimate dest<strong>in</strong>ation proves healthy. Our approach is differentfrom the secure traceroute <strong>in</strong> that 1) our approach is proposed for MANET us<strong>in</strong>g sourcerout<strong>in</strong>g protocols (e.g., DSR), the secure traceroute is ma<strong>in</strong>ly used <strong>in</strong> wired networks. 2)our approach does not require modification to exist<strong>in</strong>g rout<strong>in</strong>g <strong>in</strong>frastructures, the securetraceroute may need to modify IP layer <strong>in</strong> order to monitor performance problem; 3)our approach utilizes redundant rout<strong>in</strong>g <strong>in</strong>formation for diagnosis, the secure traceroutedoes not.Malicious nodes silently dropp<strong>in</strong>g packets exhibit the same behavior as selfish nodes,which may choose to drop packets for the sake of sav<strong>in</strong>g its own constra<strong>in</strong>t resources,such as battery or CPU cycle. Selfishness and its threat to the network performance havebeen well studied <strong>by</strong> Roughgarden [23]. Incentive mechanisms have been proposed toencourage selfish nodes to be cooperative and to forward packets for others. Unfortunately,<strong>in</strong>centive mechanisms don’t work for malicious users s<strong>in</strong>ce they never play<strong>by</strong> rules. Our proposed prob<strong>in</strong>g scheme can be used to detect and mitigate selfishnessproblem.3 Def<strong>in</strong>itions and Assumptions3.1 Node StatesWe classify the states of a node as follows. A node is GOOD if it responds to probemessages for itself and forwards other probe messages along their source routes. A

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!