A framework for secure group key management

Abstract

The need for secure group communication is increasingly evident in a wide variety of governmental, commercial, and Internet communities. Secure group key management is concerned with the methods of issuing and distributing group keys, and the management of those keys over a period of time. To provide perfect secrecy, a central group key manager (GKM) has to perform group rekeying for every join or leave request. Fast rekeying is crucial to an application's performance that has large group size, experiences frequent joins and leaves, or where the GKM is hosted by a group member. Examples of such applications are interactive military simulation, secure video and audio broadcasting, and secure peer-to-peer networks. Traditionally, the rekeying is performed periodically for the batch of requests accumulated during an inter-rekey period. The use of a logical key hierarchy (LKH) by a GKM has been introduced to provide scalable rekeying. If the GKM maintains a LKH of degree d and height h, such that the group size n ≤ dh, and the batch size is R requests, a rekeying requires the GKM to regenerate O(R × h) keys and to perform O(d × R × h) keys encryptions for the new keys distribution. The LKH approach provided a GKM rekeying cost that scales to the logarithm of the group size, however, the number of encryptions increases with increased LKH degree, LKH height, or the batch size. In this dissertation, we introduce a framework for scalable and efficient secure group key management that outperforms the original LKH approach. The framework has six components as follows. First, we present a software model for providing secure group key management that is independent of the application, the security mechanism, and the communication protocol. Second, we focus on a LKH-based GKM and introduce a secure key distribution technique, in which a rekeying requires the GKM to regenerate O( R × h) keys. Instead of encryption, we propose a novel XOR-based key distribution technique, namely XORBP, which performs an XOR operation between keys, and uses random byte patterns (BPs) to distribute the key material in the rekey message to guard against insider attacks. Our experiments show that the XORBP LKH approach substantially reduces a rekeying computation effort by more than 90%. Third, we propose two novel LKH batch rekeying protocols . The first protocol maintains a balanced LKH (B+-LKH) while the other maintains an unbalanced LKH (S-LKH). If a group experiences frequent leaves, keys are deleted form the LKH and maintaining a balanced LKH becomes crucial to the rekeying's process performance. In our experiments, the use of a B+-LKH by a GKM, compared to a S-LKH, is shown to substantially reduce the number of LKH nodes (i.e., storage), and the number of regenerated keys per a rekeying by more than 50%. Moreover, the B +-LKH performance is shown to be bounded with increased group dynamics. Fourth, we introduce a generalized rekey policy that can be used to provide periodic rekeying as well as other versatile rekeying conditions. Fifth, to support distributed group key management, we identify four distributed group-rekeying protocols between a set of peer rekey agents. Finally, we discuss a group member and a GKM's recovery after a short failure time.