NFV-Framework für sichere Übertragungen in Radio-Mesh-Netzen

Motivation and objective

Wireless data transmission represents an important means of today's network communication and is mainly found in scenarios of high mobility, ad hoc situations and scenarios with a high number of participants whose direct cabling would be uneconomical. The latter cause in particular will become much more important due to the expected increase in machine-to-machine communication and will lead to very large, so-called radio mesh networks with multihop communication. However, without adequate procedures to secure the confidentiality and integrity/authenticity as well as the availability of the communication in such networks, many application areas will not be able to sufficiently benefit from the capabilities of wireless mesh networks - in particular, existing procedures for securing wireless networks scale very poorly with respect to their key management and the planning of alternative routes to guarantee availability in case of intentional attacks.

In much scientific work on wireless mesh communications, there is a significant discrepancy between research results on the one hand and methods that can be deployed in practice on the other. In very many cases, new research concepts are only tested or evaluated on the basis of idealized modeled simulation studies [UIA12]. For example, in a broadly based literature study it was determined that only 20% of the scientific publications cite both simulation studies and experiments as evidence of their results [PKG16]. However, combining both research methods is essential for several reasons:

• On the one hand, experiments can significantly support a realistic test of presumed hypotheses while simulators often do not reflect real-world environmental parameters sufficiently accurately without extensive adjustments and parameterizations [KCR11, RLCS10].
• On the other hand, properties such as the scalability of systems can usually only be investigated reproducibly and systematically in simulators, so that simulation is also indispensable, especially for larger scenarios.

Furthermore, by using two different investigation methods, potential methodological errors in the execution of the other investigation method can be detected sooner.

Therefore, this project aims to create a framework that enables the proper evaluation of innovative protocol mechanisms for wireless communication in both the real system and the simulator. For this purpose, a component-based architecture is to be designed and developed, which will allow to considerably simplify the evaluation in the future, which is usually separated from simulations and often very complex, in a test setup with separate specifics (different operating systems, etc.). Uniform abstraction components ensure that an identical implementation can be used for both cases (simulation and real system).

Another goal is to adapt existing methods for wireless mesh networks on this basis to the special needs of users, so that they can be used in particularly safety- or security-relevant environments. In this context, the two examples "Use of 802.11ac beamforming to transmit data in mesh networks more securely against accidental and/or intentional interference" and "Implementation of hardware-independent encryption and scalable key management protocols for mesh networks" are equally singled out as use cases for the framework and as independent objects of investigation.

Approach

Parallel to the developments in wireless communications, a transition is also taking place in the wired networking world: Due to broaderband memory buses and newer programming paradigms, it is becoming increasingly possible to dispense with expensive specialized hardware and to have network services provided on the basis of conventional PC hardware, from which the communications industry and its users expect significant cost and scaling advantages in the future. In this context, the keyword Network Function Virtualization (NFV) is often mentioned. So far, however, these approaches have mainly focused on wired communication, so that support for radio hardware has not yet been adequately implemented. This gap shall be addressed in the present project.

Figure 1 illustrates the procedure of developing algorithms for wireless networks using the framework as an example of MAC, mesh routing and encryption algorithms. In each case, the implementation of these algorithms uses a uniform API, and the framework optionally provides for an implementation in simulation or real system (testbed). Thus, not only can knowledge gained during development (and especially debugging) with the help of the simulator and the prototype be incorporated into the development of new protocol mechanisms, but the implemented algorithms are simultaneously systematically tested and further developed for practical use.

For the realization of the framework, as shown in Figure 2 as an example, two versions of an abstraction layer must be developed: First, an abstraction of the used discrete event simulator and second, an abstraction of the used real hardware and operating system, respectively. Due to the good adaptation possibilities of existing simulators, the main challenges of the project are to be expected in the implementation in the real system, since, for example, the hardware API cannot be changed and further restrictions arise due to the transition between kernel and user space (memory mapping, reduction of required system calls, etc.).

A first prototype consisting of two WLAN nodes (see Figure 5) serves as the basis for the frame work development. Scaling to larger test setups is planned for the course of the project. Only standard components are used to enable a cost-efficient realization of even larger test scenarios.