Identity Access Management dataset

We release 280 synthetic IAM graphs generated using IAM graphs of commercial companies. Specifically, we vary the number of nodes, but keep graph density as is, i.e. in the range of 0.259 ± 0.198 (avg ± std). To generate a synthetic graph, we first sample the number of users and datastores from uniform distributions over the following intervals [10, 150] and [50, 300] respectively that cover variations of those parameters across real graphs. After fixing node counts we sample with replacement the actual nodes from a real world graph, which is chosen at random. Then we add Gaussian N(0, 0.01) noise to node embeddings and renormalize them. To match the graph density with the density of the underlying baseline we sample edges from a multinomial distribution, where each component is proportional to the cosine distance between a user and a datastore embeddings. Also we enforce the invariant that dynamic edges are always a subset of all permission edges. A synthetic graph generated in such a way is an ”upsampled” version of an underlying real world graph.