We consider the problem of efficiently evaluating a secret polynomial at a given public point, when the polynomial is stored on an untrusted server. The server performs the evaluation and returns a certificate, and the client can efficiently check that the evaluation is correct using some pre-computed keys. Our protocols support two important features: the polynomial itself can be encrypted on the server, and it can be dynamically updated by changing individual coefficients cheaply without redoing the entire setup. As an important application, we show how these new techniques can be used to instantiate a Dynamic Proof of Retrievability (DPoR) for arbitrary outsourced data storage that achieves low server storage size and audit complexity. Our methods rely only on linearly homomorphic encryption and pairings, and preliminary timing results indicate reasonable performance for polynomials with millions of coefficients, and efficient DPoR with for instance 1TB size databases.