A SECURE AND EFFICIENT PROTOCOL FOR ADDRESS AUTOCONFIGURATION IN MOBILE AD HOC NETWORKS

Abstract

The major research efforts in the area of Mobile Ad hoc Networks (MANETs) focus on developing efficient routing protocols which assume that the mobile nodes are configured a priori with unique addresses before they enter a network. Pre-configuration of addresses is not always possible and has some drawbacks. Furthermore, a permanent central infrastructure component or network administrator does not exist in a MANET. So, a dynamic address allocation mechanism is needed in a MANET. However, network security may be sacrificed if a reliable authentication mechanism is not present. For example, an attacker may spoof other nodes and hijack their traffic or send messages on behalf of other nodes. This may lead to Denial of service attacks. In stateless address autoconfiguration, a node, after joining an ad hoc network, assigns itself an IP address, and starts a Duplicate Address Detection phase to verify if the address assumed is not already present in the network. An attacker can now send negative replies to all the nodes wishing to join the network resulting in Denial of Service. This problem of identifying if the node is really the owner of the IP address it claims is called as the address ownership problem. In this thesis, we propose a secure and efficient protocol for address auto-configuration in a MANET based on Cryptographically Generated addresses (CGA) to solve the address ownership problem. The protocol binds each IP address with a public key which allows a node to self-authenticate itself, and thus thwarts address spoofing and other attacks associated with auto-configuration. The current IETF CGA solution is based on the RSA signature scheme. However, the cost of generating an RSA signature is high and might be prohibitive for mobile nodes which are battery powered. So, we propose to use a new CGA scheme that uses the small prime variation of the Feige-Fiat-Shamir signature scheme with Chinese Remainder Theorem and pre-computation speedup. 'Ii