next up previous
Next: Experimental Framework Up: Evaluation Methodology Previous: Software Encryption

Hardware encryption

For hardware encryption we propose to do the following
1.
Speed of encryption
Can a hardware implementation be fast enough to match the network bandwidth ?

If the software encryption does not match the network speed but the hardware encryption does, then this study becomes relevant.
Proposed table : Bandwidth vs Speed of encryption(time to encrypt one bit).

It must be noted here that the hardware implementation of a particular encryption algorithm is independent of the device on which it is present.

We will study this using available optimized hardware implementations. The feasibility of universal encryption here can easily be defined.

2.
Costs of hardware encryption (power and chip area)
How costly is hardware encryption with respect to network interfaces?

Proposed table: We analyze costs of hardware implementations with respect to standard network interfaces. This cost can be categorized as

This table will answer the question of whether it is economically feasible to incorporate an encryption co-processor into the network interface card, thereby enabling universal encryption.

In this case, the ratio of hardware costs to network interfaces costs should be very low.

A rough estimate of performance gathered from various sources [Schneier99,Goodman] is as shown in figures 1 and 2.


 
Figure 1: Speed of Software encryption, Bandwidth vs CPU
\begin{figure}
\centerline{\psfig{file=cpu.eps,width=5in,height=7.5cm}}\end{figure}


 
Figure 2: Power consumed for software encryption
\begin{figure}
\centerline{\psfig{file=power.eps,width=5in,height=7.5cm}}\end{figure}

It must be noted that the graph is only a rough approximation, not necessarily reflecting accurate values. We propose to use the above graphs as a guiding factor for our experimental framework. This framework will refine the above graphs using simulations where possible (Eg, software encryption performance). For hardware encryption, at the moment we plan to use only results obtained by researchers elsewhere.


next up previous
Next: Experimental Framework Up: Evaluation Methodology Previous: Software Encryption
Ramadass Nagarajan
2000-03-21