Two Attacks on RadioGatún

We investigate the security of the hash function design called RadioGatún in a recently proposed framework of sponge functions. We show that previously introduced symmetric trails can hardly be used to construct collisions and to find a second preimage efficiently. As a generalization of truncated differentials, trails with linear and non-linear restrictions on differences are proposed. We use these trails to find semifree-start collisions and second preimages with the meet-in-the middle approach and the complexity in the gap between claimed security level and the birthday bound. We also provide some observations on lower bounds on the complexity of our methods with respect to the length of the trail used. This is the best attack on RadioGatún.