We present an approach that we have developed in our joint project AutoSCA, as well as the prototypical integration of our approach into a product. By combining automated testing and machine learning, our method can detect side channels in TLS software to avoid serious security vulnerabilities in the future. In doing so, we analyze whether machine learning can be used to draw conclusions about the encrypted data based on the observable encrypted network traffic. This indicates a side channel and thus a potentially serious security vulnerability. Our method can be used for all TLS software, regardless of the type of implementation, hardware, or operating system since the tests and machine learning are performed at the protocol level.