Fengmin Gong, D.Sc. Manager, Advanced Networking Research MCNC 3021 Cornwallis Road Research Triangle Park, NC 27709 Email: gong@mcnc.org Phone: 919 248-9214 FAX: 919 248-1455 --white paper text-- Security and Quality of Service Integration Networking technologies have greatly enhanced our ability to gather and share information. However, the same technologies also significantly increased the vulnerability of the information to theft and distortion. The recent explosive growth in Internet applications has helped increasing the public awareness of security concerns. Nevertheless, security issues have not received the "first-class-citizen" status in the development of advanced networking capabilities. It is the author's contention that security issues must be addressed as an integral part of the Quality of Service (QoS) consideration for the development of the Next Generation Internet (NGI). Networking communities such as the Internet Engineering Task Force (IETF) and the ATM Forum are actively developing many security solutions. However, security issues are often considered as part of an after-thought process, particularly separated from other QoS development. For example, in IETF the IPSEC Working Group developed mechanisms for providing authentication and encryption services as part of the IP protocol processing, while the Integrated Services Working Group and the RSVP Working Group are developing queuing mechanisms and reservation protocol for end-to-end QoS provisioning. The security services are not included in the QoS specification. Similarly, in ATM Forum, the QoS specification has been standardized as part of the UNI and PNNI signaling protocol specifications but the security work was just started recently. There are at least three good reasons for considering security as an integral part of QoS: (1) Networking mechanisms for providing security services and those for other QoS services interact with each other; not only can degradation in one service affect the other, more efficient implementations can often be achieved by considering these mechanisms together. (2) Security concerns will be an integral part of end applications' service requirements; for example, if privacy protection is critical to an application it will not accept a connection without encryption service regardless of the bandwidth or delay characteristics. (3) In order to build survivability into the critical networking infrastructure, an adaptive security management architecture is necessary; this architecture will need to consider the end-to-end QoS as well as security requirements as a whole to make adaptive decisions for security provisioning. MCNC, through its DARPA-funded research project, has successfully developed an ATM Encryption System that operates at OC-12 (622 Mbps) rate. This stand-alone system is designed to be deployed between a private network and a public network to provide privacy protection for traffic going through the public network. The system operates in total transparency to both the public network and the end user equipment. It can support 65,534 simultaneously active virtual circuits with different encryption keys. A commercial OC-3 version of the system has been developed by Secant Network Technologies, which is a spin-off company from MCNC. Furthermore, MCNC in collaboration with MCI is implementing and experimenting with new algorithms for providing Quality of Service (QoS) guarantees over high-speed networks. Building from these projects, we plan to study the integration issues for security and QoS and to develop a survivable security management architecture for NGI. As part of this architecture, we will define the necessary specifications as well as mechanisms for providing integrated security and QoS services over NGI. We expect to be able to leverage the QoS results from the IETF Working Groups. The security management architecture will be implmemented in a prototype system and demonstrated in a testbed environment.