Malware is substantial security threat today and most likely in the foreseeable future. The analysis of malware is a key activity in the fight against the threat. Since manual analysis is time consuming and given the extent of the malware threat, malware analysis needs to be automated. Malware analysis sandboxes offer such automation and play already an important role in practice. Yet, they only uncover certain aspects of malware behavior, and still require manual analysis in many cases. This is not a viable way to go, and thus the automation and quality of automated analysis needs to be pushed further. A promising technique towards this goal is instruction tracing combined with analyzes algorithms that uncover malware behavior from an instruction trace.In this position paper, we shall argue that instruction tracing is still in its infancy and point out challenges and open problems of instruction tracing in general. In particular, we shall describe Helios, which is our new instruction tracer that offers a better balance of tracing speed and transparency than existing techniques.