Due to the globalization of the integrated circuit supply chain, untrusted third-party companies can be involved in the design and manufacturing workflow, which can bring potential security threats. Many techniques have been proposed over the years to mitigate risks and protect confidential information. However, as hardware designs become increasingly more complex, traditional design methods may exhibit reduced capabilities in handling complex, high-dimensional design spaces and usually require a substantial manual effort. On the other hand, machine learning (ML)-based methods have often been advocated in combination with model-based conventional methods as a way to harness complexity in hardware design flows. In this chapter, we survey recent approaches aiming to integrate ML techniques into a conventional hardware design flow with a focus on security objectives. We discuss their benefits and limitations and conclude by presenting potential directions for further research toward ML-enhanced hardware security.