You're using a definition of "know" that practically nobody would endorse (assuming they also accepted your other premises). Once you have certainties expressed properly as probabilities, a contextualist epistemology falls out naturally. (though there are other nuanced views that would work)