In today’s society, access to various resources (such as employment, funds, and travel) is provided only after we are able to provide evidence that uniquely identifies us. Although traditionally such evidence has been in the form of tokens, cards, and passports (things that we possess), or passwords (things that we remember), the ease of reproducing or stealing these prompted a search for physical human traits that uniquely identify who we are. Biometric measurements such as fingerprint, iris, palm geometry, palm print, face, voice, and gait are now considered as possible identification and authentication sources. Nevertheless, while most of the biometrics allow for a high level of confidence for unique identification, they also pose several problems, such as sensor variability, data variability, or simply the inability to collect the specific measurement. Combining multiple biometric measurements seems like a natural approach to improve the single biometric performance. Many factors support it, including improvements in matching, better tolerance of the inability to collect some biometrics, and increased difficulty to spoof data.

As its title suggests, this book is intended to provide an overview of the state of the art in multibiometrics research and technology. The authors are active researchers in the field, and have already published an extensive list of works closely related to this volume, giving them a strong knowledge base. The book is part of a larger international series on biometrics research of which one the authors of this book (Jain) is a consulting editor. A browse through other titles in the series reveals topics such as face, gait, and fingerprint recognition. In this context, a handbook that deals with multiple biometric measures can be seen as a capstone of the series and a must-have for any professional in the field.

The volume is organized in a clear fashion, and has five main chapters that deal with general biometrics, information fusion, levels of data fusion, fusion using scores, and other factors in biometrics fusion, respectively. For a reader interested in a short review of the current state of the art in biometric identification, the first chapter is a must. In it, the authors provide descriptions of all the major biometric measures, discuss accuracy issues, and list several currently deployed biometrics systems. A significant drawback of this introduction is that the text does not address any issues related to individual privacy and confidentiality of the biometric data, factors certain to increase in importance as more systems are deployed. For computing and information technology professionals, it is also interesting to note that, among other factors potentially affecting authentication, the authors include general information security issues such as data corruption or transmission interception and processing module replacement. The existence of such factors suggests that, as with any security problem, one must ensure protection at all levels, not just at the biometric one.

The second chapter introduces the reader to general information fusion, and then focuses on fusion of biometric data. Two approaches are summarized: sensor level (mostly applicable to the same biometric) or decision level (where one biometric match result is used together with other biometric results to indicate if further biometric matches are required). The third and fourth chapters discuss various fusion approaches.

The final chapter focuses on the use of additional information in the decision system. One such piece of information is the quality assessment of the collected data. In this case, noisy data (due to lack of maintenance of the sensor or to poor collection environment conditions) affects the results and should be included in the decision (providing an additional confidence measure for the accuracy). Another approach is the use of soft biometrics measures (such as gender, ethnicity, height, weight, and eye color) in improving the decision. The justification for such measures is that many biometric data systems include this information, and author-originated studies suggest that using it increases accuracy. While the authors discuss the technical advantages and disadvantages of using such measures, the chapter lacks any discussion of their subjectivity or any of the ethical implications that may arise.

Overall, the book is easy to read and includes a significant body of background literature. The authors also include a list of currently available multibiometric databases, as well as suggestions for the evaluation of multibiometric systems. The text, while targeted to biometrics professionals, could also constitute a worthy lecture for anyone interested in imaging, computer vision, or information fusion. With some skipping over of the mathematical formulations, any computing professional can benefit from reading it, as the book constitutes a good example of the current research in the biometric authentication field.