 This talk will be about a new idealized model for hash functions called the augmented random oracle model. So imagine we have some cryptosystem that makes use of a hash function H. Unfortunately, no matter how hard we try, we cannot prove security based on typical properties of H, such as collision resistance. Nevertheless, the scheme seems to resist attacks. So what do we do? One proposal by Bilare and Ragoe is the famous random oracle model or ROM. Here, the hash function is modeled as a uniformly random function O that can only be accessed via queries. The idea is to prove security in this idealized model. This gives at least some assurances that we didn't totally mess up the cryptosystem design. Then we just pray that security holds when we replace the random oracle O with a concrete hash function, say SHA2. Unfortunately, Kaneti Goldreich and Halevi showed that random oracles must sometimes fail. They give what is called an unistantiability result for the random oracle model by exhibiting a scheme S that is secure in the random oracle model but insecure under any concrete hash function. Since their work, numerous other unistantiability results have been shown. Despite these results, the ROM remains widely used and is a very important tool for analyzing practical cryptosystems. In many cases, the random oracle model is the only way we know how to justify the security of practical cryptosystems. So the question is, what can we do to get a better understanding of the guarantees of the ROM despite these problems? This brings us to the goal of our work, which is to refine the random oracle model so as to avoid unistantiability results. We do exactly this by defining a model that captures all known techniques used to give unistantiability results while still maintaining some of the utility of an idealized model. Please watch the talk for how we design such a model and what we can prove.