On a Relation between Knowledge-of-Exponent Assumptions and the DLog vs. CDH Question