Key Management for Dynamic Peer Groups in Mobile Ad Hoc Networks

Introduction

Mobile ad hoc networks (MANETs) allow users to establish communication without any fixed or pre-existing infrastructure. The network therefore has no base stations, access points or remote servers. Nodes that are within each others transmission range communicate directly, while relaying the messages for those too far apart. The mobility of the nodes can lead to ‘rapidly changing’ (dynamic) network topologies. Nodes do not have any relationships prior to network formation due to the nature of the applications of MANETs (van der Merwe et al., 2007).

Generic MANETs are created solely by the end-users for a common purpose in an unplanned, i.e. ad hoc fashion. In contrast to conventional networks the users therefore cannot bootstrap the required security associations with the assistance of a priori shared information on their nodes. This unique property demands distributed collaborative protocols that enable nodes to establish security mechanisms without the assistance of a centralized online Trusted Third Party (TTP).

Many researchers have already proposed peer-to-peer key management schemes that are suitable for self-organized and authority-based MANETs (van der Merwe et al., 2007). In contrast, the available literature contains very few group key management schemes that are designed specifically for MANETs. Authors normally adapt group key management schemes for conventional networks to suite the unique characteristics of MANETs.

The military and commercial applications of MANETs incorporate many group-oriented applications (van der Merwe et al., 2007). The primary subdivision of group key management schemes emerges from the variety of different application dependent group settings that exist in practice. In Figure 1, the two main group key management settings are indicated within the dotted lines, namely:

Figure 1.

Dimensions of group key management

• •

  Group key management for centrally managed, non-collaborative groups

• •

  Group key management for dynamic peer groups (DPGs)

Large groups, found for example in internet multicast applications, are normally non-collaborative and hard to control on a peer basis (Steiner et al., 2000). They therefore have a structured hierarchy and exhibit one-to-many broadcast communication patterns (Steiner et al., 2000; Kim et al., 2004). The control structure is maintained by a centralized TTP, chosen prior to network formation.

DPGs tend to be relatively small collaborative groups (with membership in the order of a hundred) where all group participants have a symmetric relationship and must therefore be treated equally (Steiner et al., 2000). Such systems accordingly have no central point of control. This means that special roles, such as a group controller, are also not fixed prior to group formation, but allocated to any group member during and after group formation. These roles must be assigned based on group policy and must be orthogonal to the key management scheme (Steiner et al., 2000).

DPGs have many-to-many communication patterns and are dynamic in membership, i.e. members join and leave at random. Since a common cryptographic key must be shared between group members at all times, the dynamic membership makes key management protocol design complex as the computational and communication overhead on the network has to be kept to a minimum (Kim et al., 2004). Dynamic membership result in group key agreement protocol suites that accommodate initial key agreement (IKA) and auxiliary key agreement (AKA) operations (Steiner et al., 2000). IKA refers to the key agreement during the first group formation or group genesis, while AKA includes all subsequent key agreement operations. The most common AKA operations are illustrated in Figure 2 (Steiner et al., 2000).

Figure 2.

Common auxiliary key agreement operations