Building an Application-Aware IPsec Policy System

As a security mechanism at the network-layer, the IP security protocol (IPsec) has been available for years, but its usage is limited to virtual private networks (VPNs). The end-to-end security services provided by IPsec have not been widely used. To bring the IPsec services into wide usage, a standard IPsec API is a potential solution. However, the realization of a user-friendly IPsec API involves many modifications on the current IPsec and Internet key exchange (IKE) implementations. An alternative approach is to configure application-specific IPsec policies, but the current IPsec policy system lacks the knowledge of the context of applications running at upper layers, making it infeasible to configure application-specific policies in practice. In this paper, we propose an application-aware IPsec policy system on the existing IPsec/IKE infrastructure, in which a socket monitor running in the application context reports the socket activities to the application policy engine. In turn, the engine translates the application policies into the underlying security policies, and then writes them into the IPsec security policy database (SPD) via the existing IPsec policy management interface. We implement a prototype in Linux (Kernel 2.6) and evaluate it in our testbed. The experimental results show that the overhead of policy translation is insignificant, and the overall system performance of the enhanced IPsec is comparable to those of security mechanisms at upper layers. Configured with the application-aware IPsec policies, both secured applications at upper layers and legacy applications can transparently obtain IP security enhancements.