You need to learn how the SNP values are named and what they mean. The short answer is that the SNP is particular base in the DNA. It has it's own "address" from one of the ends of DNA helix. SNP's value is the aminoacid located in this position. We made so-called reference genome for convenience. It is described as chain of SNPs with strictly fixed value corresponding to one particular common ancestor of all mankind. The selection of this reference genome is relative. When we will have more data we can change the SNPs value in it to correlate with new information and new conditions.
So in this view the "ancestral" term becomes clear. It means that this particular address in DNA contains the same value as in this reference DNA. If one of your ancestors had the mutation the SNP value will be different because the different aminoacid will be found in this place. It is "derived" SNP.
Unfortunately, sometimes very rare event happens: when back-mutation occurs. It means that the SNP changes it state from "derived" to "ancestral" and all your issue will have again the ancestral version. When we build BIG tree of relationships we try to analyze whole dataset to minimize this effect.
The simple answer also can be obtained via the link:
Ancestral = unmutated version of an SNP
Derived = mutated version of an SNP