Catch Packet Droppers in Wireless Ad-hoc Networks Wei-Cheng Xiao, Lei Tang Mobile Ad-hoc Network • • • No access point, self-configured Every node is a router. Packet transmission relies on node cooperation. 2 Motivation Catch packet droppers in wireless ad-hoc networks that: 1) Maliciously drop packets 2) Drop packets due to network partition or poor link quality 3 Challenges and Assumptions 1. Challenges • • No global monitoring No trust relationship among nodes 2. Assumptions • • Source-routing based packet transmission Power of dropper • • • Random drop only Cannot modify packet content Pre-distributed pairwise shared secret key. 4 The Proof-based Scheme Packet Packet S A Proofhop D B Proofe2e Proofhop = {src, dest, seq, I, HMACKS,B(src, dest, seq, I )} Proofe2e = {src, dest, seq, HMACKS,D(src, dest, seq)} • • • • Node X is unable to forge proofs between node Y and Z. The source is able to verify the proofs. Only symmetric cryptography (HMAC-SHA1) is used. Intermediate nodes cache proofs for later investigation. 5 Investigation The source requests each node on the route to present its cached proofs. Investigation Proof 6 Aggregated Investigation To reduce overhead, multiple packets are investigated using only one investigation and one reply. Investigation: source destination start seq. Reply: end seq. Hash(proofs) Bitmap of packets having a proof 7 Dropper detection and Dropping Metric packet 1 2 refuse proofhop packet 3 • Narrow down the scope of possible droppers via proof analysis • Dropping metrics of nodes Maintained by each source, not shared with other sources. • Metrics computation M (1 ) M Inc. Possible dropper: metrics increased Other nodes: metric decreased (has minimum metric) • Periodical metric decreasing for nodes not on the routes Give chance to nodes that are no longer bad 8 Evaluation • • • We implemented CATCH in ns2 network simulator. Parameter Inc= -1 when a packet is delivered to the destination. Inc=3, when a node is unable to present a proof. α is set as 0.2. In the experiments, DSR was used as the routing protocol. 9 10-node-chain, node 5 drop rate 1.0 10 10-node-chain, node 5 drop rate 0.2 11 49-node grid network—false negative rate 12 49-node grid network--false positive rate 13 Conclusion • With CATCH, a dropper ends up having a higher dropping metric than non-droppers. But false positives are inevitable. • In most experiments, CATCH scheme successfully caught the droppers. 14