Learning a Neural-network-based Representation for Open Set Recognition
This addresses the challenge of handling emerging unknown classes in domains like security, such as malware classification, but appears incremental as it builds on existing representation methods.
The paper tackles the problem of open set recognition, where systems must identify unknown classes in addition to discriminating known ones, by proposing a neural-network-based representation that clusters same-class instances and separates different-class ones, resulting in statistically significant improvements on three datasets from two domains.
Open set recognition problems exist in many domains. For example in security, new malware classes emerge regularly; therefore malware classification systems need to identify instances from unknown classes in addition to discriminating between known classes. In this paper we present a neural network based representation for addressing the open set recognition problem. In this representation instances from the same class are close to each other while instances from different classes are further apart, resulting in statistically significant improvement when compared to other approaches on three datasets from two different domains.