Title: Data-driven Approach to Android App Outlier Analysis Abstract In my talk, I will present our recent study on outliers in Android apps. Consider a group of apps that require permissions to use cameras. By analyzing public information, we can discover these apps tend to mention relevant words such as camera. By analyzing their user interfaces, we can discover these apps tend to include relevant buttons such as a “take picture” button. By analyzing their code, we can discover these apps tend to invoke relevant camera API methods. A model can be trained to capture these common patterns. An outlier is the one that does not fit the model. Why does this app ask for the camera permission but never mention a word about it? Is it trying to hide something? Why does this app ask for the camera permission but never invoke any API method to operate the camera? Is this a bug, malice, or a feature? We take a data-driven approach to analyzing Android apps for uncovering outliers. We compute app descriptors by extracting features at the public level, the interface level, and the code level. I will report our findings on a large corpus of 178,765 apps with respect to camera permissions, location permissions, phone permissions, permission count, advertisements, content ratings, categories, install size, and near identical duplicates.