The influenza virus is known for its rapid evolution, or the ability to fix many mutations over a short period of time. Some of these mutations lead to amino acid substitutions in regions of the virus targeted by the immune system. Such changes are often selected for because they confer a fitness advantage by allowing the virus to "escape" immune response. This pattern of repeated immune escape is a detriment to public health because it necessitates an annual update to the influenza vaccine. Therefore, identifying sites on the influenza virus which are targeted by the immune system could help predict which influenza strain will circulate in the future, inform vaccine design, and help understand basic evolutionary questions. Using molecular phylogenetic techniques, we can identify sites potentially targeted by the immune system by looking for "positive selection", a phenomenon which manifests as a higher than expected rate of evolution. To identify sites evolving faster than expected, we defined a null expectation of the evolutionary rate of influenza in the absence of immune pressure. This null model is defined using empirical measurements from a high-throughput functional assay known as deep mutational scanning. This null model differs from traditional phylogenetic models in that it describes the constraints on influenza on a site-specific basis and, as a result, has been shown to be a more accurate and powerful null model. I have implemented an empirical Bayes approach to identify sites which deviate from the null model by an unexpectedly high evolutionary rate, suggesting positive selection. Preliminary results show that my method outperforms other methods for identifying sites under positive selection. Next, I will apply these methods to the influenza virus surface protein, hemagglutinin, which is a major target of the immune system.