Analyzing Memorization in Large Language Models through the Lens of Model Attribution