Updated the publications list to include the fruits of the collaborations with RECOD (Unicamp) on forensic applications of Machine Learning.
The two papers provide simple, yet practical solutions in a domain where the accountability and interpretability of the results are of paramount importance. Both papers were submitted to peer-reviewed journals with good impact factor (unfortunately, some funding agencies prefer this approach to the usual conference standards).
The papers are also interesting in that they help set the stage so that latent variable probabilistic models (including, for example, Gaussian Processes and Variational Autoencoders) can be used together with causality for an analysis of which features of forensic evidence lead to attribution to a given class of interest (e.g. if a request for a set of permissions causes a malware modus operandi, or, for example, if a smudge in a printer image implies it was produced by a given printer).
There are many interesting avenues to follow from this, and it was great to work with them last year. They are very serious applied researchers (and also really hard working)!