Network-Aware Security for Group Communications

In these scenarios, group members subscribe to different data steams, or possibly multiple of them. In other words, the access control mechanism needs to supports multi-level access privilege. This is referred to as the hierarchical group access control [115, 116].

Many existing key management schemes focus on maintaining key secrecy and reducing the communication overhead associated with updating the associated keys [78] [7] [56]. However, it is found that key management can disclose information about dynamic group membership to both insiders and outsiders. In other words, while the content of group communication is protected by encryption using the secret keys, group dynamic information is disclosed through key management. Group dynamic information (GDI) is the information that describes the dynamic group membership, including the number of users in a multicast group as a function of time, and the number of joining or departing users in a time interval.

In many secure group applications, group dynamic information should be kept confidential [123, 124]. Key management is a technology that enables key updating in real time as group membership changes. Future commercial multicast services, which could occur in non-traditional broadcast media such as Internet and 3G/4G wireless networks, will allow a user to subscribe to an arbitrary set of programs and change his/her subscription at any time [125] [10]. The users can choose to pay for exactly what they get, instead of a fixed monthly fee. This new type of services give the most flexibility to users, as well as opportunities to new business models. Over the non-traditional broadcast media, the global media giants as well as small multimedia producers can be the service providers. The service providers perform group management and have the knowledge of GDI, i.e audience statistics. However, it is highly undesirable to disclose instant and detailed GDI to competitors. Assume a competitor can monitor the audience statistics of the service provider X. Then, the competitor may broadcast its programs at different time slots and see how it affects its own and X’s audience statistics. As a consequence, the competitor can develop the best program schedule to compete with X. This example also shows that GDI should also be concealed from insiders. A regular user, who receives the multicast content, should not know the overall audience statistics. Otherwise, the competitor can send one of its employees to register as X’s member for a small cost, and collect valuable audience statistics from X. In addition, there are multicast communication scenarios where GDI represents sensitive deployment information about the network. For example, in a sensor network, the base station sends many broadcast messages to sensors. The base station and sensors form a secure multicast group. If some sensors are compromised, the group key should be updated such that the compromised sensors cannot decrypt future multicast messages from the BS. One possible way to update group keys is to use group key management schemes. In such an application scenario, GDI represents the number of sensors deployed in an area, and the number of revoked sensors. In this example, if GDI is not protected, attackers can obtain sensor deployment information by exploiting the key management scheme.

In this chapter, our objective is to present strategies that reduce the delay associated with multicast authentication, make more efficient usage of receiver-side buffers, make delayed key disclosure authentication more resilient to buffer overflow denial of service attacks, and allow for multiple levels of trust in authentication. Throughout this chapter, we will focus our discussion on the popular multicast authentication scheme, Timed Efficient Stream Loss Tolerant Authentication (TESLA), though our techniques can apply to other authentication methods based upon the delayed key disclosure principle. Like other schemes based upon delayed key disclosure, TESLA is susceptible to DoS attacks and is not well-suited for delaysensitive applications. At the heart of our approach is a modification to TESLA, which we call Staggered TESLA, that employs several message authentication codes (MACs) that correspond to authentication keys that are staggered in time. Staggered MACs provide notions of partial authentication and allows for forged packets to be more readily removed from the buffer, thereby improving usage of the receiver’s buffer. A benefit of partial authentication is that one may define security policies that allow for partially authenticated packets to pass through the buffer, and thus packets will remain in the buffer for a shorter duration. In many scenarios accepting partially authenticated packets is unacceptable, and therefore we present two further techniques that may be used to reduce the delay needed for full authentication. The first strategy requires that the source has a guarantee that there are no adversaries within a certain network distance of the source. By having a guarantee of proximity protection, partially authenticated packets may be accepted as fully authentic. The second strategy for reducing full authentication delay that we present involves replicating the key distribution functionality within the network, and having a set of distributed key distributors transmit the key seeds. A benefit of all of these strategies is that they mitigate the threat of a buffer overflow DoS attack since an adversary must conduct a DoS attack at a higher attack rate.