CHEF: A Configurable Hardware Trojan Evaluation Framework

Evaluating approaches to hardware Trojan detection often involves the use of benchmark hardware designs in order to maintain comparability and reproducibility of experiments. However, such benchmark designs may have to be adjusted to the tool chain that is used for evaluation. Also the tool chain may have to be adjusted to an individual benchmark design. Such adjustments can be time consuming, particularly if these adjustments have to be made multiple times. In order to maintain reproducibility, such adjustments have to be documented in a way that an experiment can be repeated. It is also important that the inclusion of malicious functionality is clearly documented. In this work, we present a methodology to clearly document changes to benchmark designs and tool chains. We facilitate the advantages of the extensible markup language (XML) in order to create a set of rules that is applied to the original benchmark design or tool chain which then yields the desired adjusted benchmark or tool chain. Additionally, we provide facilities to describe an experiment in a structured way, which allows to automatically perform such an experiment. Also, it is possible to share an experiment description such that an experiment can be repeated by peers. That way, the derived conclusions from such an experiment can be dramatically strengthened.