Neural networks are modeled after the structure of the human brain, and the human brain has an advantage over a computer in terms of pattern recognition. It is likely, therefore, that neural networks might be suitable for identifying striations such as toolmarks and impressions on fired bullets. Generally, there are two types of neural networks: a Hopfield network and a multiplayer network (MLN). This study used the MLN, which has a structure of several layers, with each layer consisting of a number of nodes called neurons. Each neuron in a layer connects to all neurons of the next layer, and no one neuron connects to any other neurons in the same layer. The model used in the current study contained three layers, with the bottom layer being the input layer and the top layer the output layer. This paper reports on the retrieval of actual bullets from 10 firearms (9 mm Ruger) by using the MLN. The MLN used in this study deals with binary signals derived from striation images. The signal has a significant role in identification because it is the key to the individuality of the striations. The neural network searches a database for similar striations by means of these binary signals. The algorithm developed in this study was effective in identifying bullets and was able to identify deformed striations. It was not necessary to adjust strictly the location of bullets obtained from an image of a land impression; this could significantly shorten the time needed for an inquiry. A much larger number of bullets will be required for the validation of the algorithm. If the database contains more bullets, there is no guarantee that the MLN will be able to select the right match. The aim of such an automated comparison system is to reduce the labor time of a scientist; however, the decision on a bullet identification must finally be made by a qualified forensic scientist. 1 table, 9 figures, and 5 references