The present study aims to contribute to three aspects in iris recognition technology. First, in the area of database search, an alternative search technique that has been used in operational 1-to-many matching scenarios was evaluated. Results show that critical degradation of the error rates can occur when compared to the traditional approach. Further evaluation showed, however, that upon careful selection of operational parameters it is possible to obtain reduction in the number of comparisons performed, with minimal loss in accuracy. Iris presentation attack detection (PAD) is the second area approached by this work. Recent works show that the ability to identify artifacts in cross-domain scenarios is still limited. This work proposes a new method for PAD that is based on CNN classification of multiple views of the data, followed by fusion of the multiple results. Evaluation shows the new approach outperforms the current state of the art in cross-domain PAD. Finally, gender prediction based on iris images is the last domain explored. Our study shows that despite an optimistic trend in related works, the actual potential of the iris texture for gender discrimination is very limited, and most of the information that is used in the process comes from surrounding regions.