Strong physical unclonable function (PUF) is a promising solution for device authentication in resource-constrained applications but vulnerable to machine learning (ML) attacks. In order to resist attack, many defenses have been proposed in recent years. However, these defenses incur high hardware overhead, degenerate reliability and are inefficient against advanced ML attacks such as approximation attacks. To address these issues, we propose a <underline>R</underline>andom <underline>S</underline>et-based <underline>O</underline>bfuscation (RSO) for Strong PUFs to resist ML attacks. The basic idea is that several stable responses are derived from the PUF itself and pre-stored as the <inline-formula> <tex-math notation="LaTeX">$set$ </tex-math></inline-formula> for obfuscation in the testing phase, and then a true random number generator is used to select any two keys to obfuscate challenges and responses with XOR operations. When the number of challenge-response pairs (CRPs) collected by the attacker exceeds the given threshold, the <inline-formula> <tex-math notation="LaTeX">$set$ </tex-math></inline-formula> will be updated immediately. In this way, ML attacks can be prevented with extremely low hardware overhead. Experimental results show that for a <inline-formula> <tex-math notation="LaTeX">$64\times 64$ </tex-math></inline-formula> Arbiter PUF, when the size of <inline-formula> <tex-math notation="LaTeX">$set$ </tex-math></inline-formula> is 32 and even if 1 million CRPs are collected by attackers, the prediction accuracies of the several ML attacks we use are about 50% which is equivalent to the random guessing.