Protecting privacy with protocol stack virtualization

Previously proposed host-based privacy protection mechanisms use pseudorandom or disposable identifiers on some or all layers of the protocol stack. These approaches either require changes to all hosts participating in the communication or do not provide privacy for the whole protocol stack or the system. Building on previous work, we propose a relatively simple approach: protocol stack virtualization. The key idea is to provide isolation for traffic sent to the network. The granularity of the isolation can be, for example, flow or process based. With process based granularity, every application uses a distinct identifier space on all layers of the protocol stack. This approach does not need any infrastructure support from the network and requires only minor changes to the single host that implements the privacy protection mechanism. To show that no changes to typical applications are required, we implemented the protocol stack virtualization as a user space daemon and tested it with various legacy applications.