If you mean words that come from a single word, you can call them cognates. If you mean words that sound similar, phonological neighbours. There is no single unit of measurement for how similar two words are, but the most straightforward and common way (though by no means necessarily the best) is to do this:
Decide on a way of measuring distance between phone(mes) in the two languages, e.g. /p/ and /b/ are probably more similar than /p/ and /h/
Use a way of calculating edit distance, most commonly the Levenshtein distance, to calculate the distance between the two strings, using the distance in the first step as the cost of a substitution operation, and some ratio (say, half) of the average substitution cost as the addition or deletion cost.
I've gone into much greater detail on this Reddit post, which you can read if you're interested.
Your examples show cognates, i.e., words that are etymologically related. Cognates is used here in the wider sense including borrowed words.
Given a list of established cognates, it is possible to derive reconstructed protoforms and phonological rules ti derive the respective forms. Once known, you can try to identitfy more cognates given the rules. Roman Yangarber did something like this (with more than two languages involved) for the Finno-Ugric languages.