Adaptive E-learning with Transparent User Identification Kamen Kanev, Noriaki Kamiya, Nikolay Mirenkov, Go Kanda, Akiyuki Takagi University of Aizu, Aizu-wakamatsu City, Fukishima-ken, 965-8580, Japan {kanev|kamiya|nikmir}@u-aizu.ac.jp Abstract In this work we discuss methods and technologies for transparent user authentication in the context of adaptive e-learning. A new collaborative elearning environment model with support for adaptive applications is proposed. Based on it an augmented interactive desktop that allows individual and shared use of digitally enhanced educational handouts as tangible interface components is developed. Cluster Pattern Interface (CLUSPI) enabled cards for elementary math tutoring are designed and educational activities at individual, group, and class levels are considered. Keywords: E-learning, collaborative learning, transparent authentication, RFID, fingerprint recognition, adaptive applications. 1. Introduction User recognition and authentication play an important role in e-learning, especially when adaptive applications are considered. Widely used password-based authentication, however, appears to be difficult to employ when quick, transparent context changes are required. Frequent user changes that naturally occur when two or more learners are collaborating and sharing an elearning workstation, for example, can hardly be handled through password authentication. Furthermore, in adaptive applications, user context management and personalization of application interfaces and functionality have to be conducted on a continuous basis. In our previous studies [1] we have investigated collaborative language learning in a dynamic group environment where students are free to work individually, to do pair work, or to form larger groups at any time. When learning, students could use digitally enhanced printed materials and supportive digital content, accessible through shared CLUSPI readers [2]. For gathering proper feedback about group formation, interaction in groups, and group disbanding we have suggested employing RFID tags assigned to students and compact handheld RFID readers integrated with CLUSPI devices. In this work, along with the continued discussion of RFID-based authentication, we also consider biometric methods for user identification and authentication [3]. Since in our experiments we work with elementary school children, we believe that the solution eliminating the need of digitally enhanced user IDs and using purely biometric data might be more suitable [4]. 2. Methods and technological means for transparent user authentication Transparent user authentication refers to recognizing people on the fly, without disturbing or interfering with their current activities. Such identification happens, for example, in quite a natural way when people interact between themselves. In human societies, we can quickly recognize each other by the way we look, by the way we speak, and by the many other distinctive features that make us unique. For a computer system, however, recognition and authentication of humans on the fly is far from a trivial task. Consequently, instead of addressing this problem, many of the commercial user authentication systems would rather rely on digitally enhanced user IDs. Those could be in the form of badges with barcodes, magnetic stripes, embedded RFID (Radio Frequency Identification) tags, etc. RFID-based methods seem to be particularly suitable for transparent user identification, since they provide instant access to ID data without contact or clear line of sight requirements [5]. Our RFID-related studies and experimental work are reported in the following sections. A potentially better solution for transparent user identification would be to eliminate the need of digitally enhanced user IDs and to use purely biometric data [3,4]. Here we present a nonexhaustive list of currently available biometric authentications methods that could be presumably employed for transparent user identification: Face Recognition Voice Recognition Signature Recognition Iris Recognition Retina Recognition Fingerprint Recognition Hand Recognition Finger Geometry Palm Geometry People are good at recognizing human faces and voices, but unfortunately currently available face and voice recognition methods and algorithms lag far behind, and remain of limited practical use. Other enlisted biometric methods such as iris, retina, and signature recognition require voluntary user actions and thus are not suitable for transparent user authentication. In human-computer interaction, the two most commonly used input devices are the keyboard and the mouse. During the interaction process, humans are constantly in contact with these input devices. Therefore, for a truly transparent user authentication, we would only need to make the computer sense and recognize human fingers or hands in the course of the input process. We thus mark the hand and finger related biometric methods on the above list as good potential candidates for more detailed studies and experiments. Further discussions and details about appropriate input devices with built-in biometric sensors, suitable for transparent user authentication will be given in the following sections. The proper selection of RFID tag types, their sensing ranges and the way users wear them thus becomes a very important issue. Limited range hand-wearable RFID tags, for example, would be most suitable for properly tracking the person currently in charge of the input. Wider range RFID badges, on the other hand, would suit for identifying members of a group, gathered in front of a computer monitor. The method that we chose to employ is based on wearable RFID rings and on compact RFID readers, integrated with traditional input devices such as keyboards, mice, handheld scanners, and CLUSPI readers. For illustration we show in Fig.1(a) several such rings created by attaching button-type RFID tags to magic tape stripes and conventional plastic rings. (a) (b) Figure 1. RFID rings (a) and a RFID mouse (b). The RFID rings can be used when operating RFID-enabled keyboards as discussed in [7] or the RFID mouse shown in Fig.1(b). 2.1. RFID-based user authentication Currently available RFID technologies work with passive tags of different shapes and sizes that can be sensed from a few centimeters to a few meters and with active, battery powered tags that can operate at even larger distances [6]. In contrast to optical technologies like barcoding, where a clear line of sight is always required, the RFID tags can be accessed even in presence of visual obstacles. RFID technologies are also more stable and prone to damages and errors, and do not require contact or close proximity as magnetic stripes. In the context of e-learning, truly transparent user identification and tracking of user activities could only be achieved if users do not have to take any conscious actions to authenticate themselves. Figure 2. The combined RFID-CLUSPI reader. For our experiments we have also prepared an integrated RFID-CLUSPI reader that combines the CLUSPI direct point-and-click functionality [8] with the RFID user identification. For illustration, a user wearing one of the rings and holding the developed RFID-CLUSPI reader is shown in Fig.2. User identification is conducted on the fly in a transparent and unobtrusive way, whenever users interact with the RFID-enabled device. In this way collaborative groups of learners, sharing existing e-learning environments, could be created and managed. 2.2. Fingerprint-based user authentication RFID-based authentication methods discussed in the previous section require all users to wear RFID tags which may cause some inconveniences, especially when children and elderly people are involved [4]. In this context, alternative approaches, such as fingerprint identification, vascular pattern recognition and other biometrics authentication methods that do not require tagging are worth considering. Fingerprint authentication is a mature technology and affordable development systems and software kits are available from different makers. We will, therefore, continue with a more detailed discussion of some current fingerprint authentication equipment and supporting software development tools. becomes an alternative, or a complete replacement of the standard password-based user authentication embedded in the operating system. While this is a convenient approach that establishes fingerprintbased control over the user access to system facilities, software and data, it does not allow more granular, application level access management and control. For access to such lower level functionality appropriate SDK and tools, which are provided only at the discretion of the fingerprint reader producer, are necessary. Based on this criterion we have acquired from RATOC Systems, Inc. the fingerprint reader hardware, software and development tool set shown in Fig.4. Figure 4. The fingerprint reader hardware, software and tools from RATOC Systems, Inc. The currently available SDK supports only Windows-based development and thus for now we conduct all our experimental work on Microsoft Windows machines. (FAM) Fingerprint Authentication Module Input of Fingerprint Images Feature Extraction User Database Search Registered User Database Authenticated User ID (FRD) Fingerprint Reader Device Figure 3. Fingerprint authentication system. A typical fingerprint authentication system (Fig.3) consists of a fingerprint reader (1) connected to a computer (2) and fingerprint authentication software (3) that integrates with the operating system of the employed computer. In this way, fingerprint-based user authentication E-learning Application (UMM) User Management Module User Profile Data Queries User ID Input Profile Data Request Profile Data Input Figure 5. Structure of the developed fingerprint identification software. The structure of the developed fingerprint identification software and its integration with separately built applications is shown in Fig.5. The Fingerprint Authentication Module (FAM) first obtains fingerprint images from the fingerprint reader through a USB connection. Feature extraction algorithms are then applied to the acquired images and, based on the result, a search in the user database is conducted. If a positive user authentication is obtained, the corresponding user ID is sent to a separately built application directly or via the User Management Module (UMM) as depicted in Fig.5. Note that the FAM module operates in a continuous mode and produces a stream of authenticated user IDs as long as a properly registered user finger stays within the sensing range of the fingerprint reader device. This functionality is a basis of the core transparent user authentication that we propose and allows separately built applications to get notifications of user changes in real time. Some fingerprint reader devices, however, appear to lack continuous fingerprint input capabilities. The device shown in Fig.6, for example, requires a sliding finger gesture to be performed for every fingerprint input. Obtaining a stream of authenticated user IDs is then only possible if the user constantly slides his finger over the sensor, which obviously does not constitute an acceptable interaction method. Figure 6. A sliding fingerprints reader. To implement truly transparent user authentication one would actually need a fingerprint reader device integrated with other traditional input devices in such a way that users do not even notice its presence. The fingerprint mouse shown in Fig.7 is one such example and we are considering using it for comparative studies during our experimental work. When a user holds the fingerprint mouse his thumb stays within the sensing range of the embedded fingerprint reader and thus allows continuous sensing and streaming of fingerprint images. Once the user releases the mouse, however, his thumb leaves the fingerprint reader sensing range and the streaming of fingerprint images is interrupted. In this way multiple users could share a single mouse and still the current user would be always positively identified. Figure 7. A fingerprint mouse. There are also fingerprint keyboards that employ an embedded fingerprint reader for user authentication. Unfortunately, with such keyboards, fingerprint scanning and typing are two independent processes that cannot be easily integrated. It is not difficult to envisage, however, that future keyboards might have contact fingerprint scanners embedded in all their keys and then, simultaneous typing and authentication would become possible to implement. Similarly to the fingerprint mouse in Fig.7 we are considering an integrated fingerprint-CLUSPI input device. It could be constructed by combining a compact fingerprint reader with a standard CLUSPI device in a way much similar to the one employed for building the integrated RFIDCLUSPI device shown in Fig.2. The resulting fingerprint-CLUSPI reader, although being a single integrated device, could still function as two independent devices with separate USB connections. In this way no special software for the integrated device would be need and the standard drivers and previously developed CLUSPI related and fingerprint authentication based applications could be used with no modifications. 3. Digitally-enhanced collaborative e-learning environments We are currently developing a new collaborative e-learning environment model with extended support for adaptive applications and transparent user identification. Based on it, an augmented interactive desktop that allows shared use of digitally enhanced physical objects as tangible interface components, is being constructed [9]. The augmented interactive desktop could be used in practical experiments with digitally enhanced printed handouts for elementary math tutoring and language learning [10]. A schematic diagram of the currently developed software system is shown in Fig.8. User Database (FRD) Fingerprint Reader Device Additional Authentication Device(s) (RAM) RFID Authentication Module (FAM) Fingerprint Authentication Module Additional Authentication Module(s) FAM Section Other Section(s) Authenticated User ID (UMM) User Management Module User ID Interface RAM Section User Profile Data (RRD) RFID Reader Device User Profile Data Queries User Profile Access Interface E-learning Application E-learning Application Figure 8. A schematic diagram of the developed software system. The system has a modular structure and is specifically designed to support different user authentication methods. At present we have a fully functional prototype of the RFID authentication module while the fingerprint authentication module is still under implementation. Additional user identification modules will also be developed and added to the system in the future. Since different authentication methods could employ different input devices, a default authentication method is supplied to the system during the initialization. Based on that, an appropriate user authentication module is selected and access to the corresponding user database sections is ensured. When a user is authenticated by any of the available authentication modules its unique user ID is sent to the UMM module as explained in the previous chapters. Here we will continue with an outline of the UMM module functionality. The UMM module is mainly responsible for providing independently developed applications with user related data, extracted from the user database according to their needs. One application may need to know, for example, the age of the users while another may need their genders. Instead of allowing every application to query the user database independently, one could gather and process all the queries within the UMM module and thus ensure more efficient and secure database access. Following this model, applications could register the user profile parameters of interest with the UMM User Profile Access Interface. Afterwards, whenever a user is successfully authenticated, the UMM will be automatically providing the required user data to the relevant applications. 4. Experimental framework elementary math tutoring for adaptive In this section we discuss some practical aspects of the collaborative e-learning environment model presented earlier. Our objective here is to build an experimental framework, based on the model and employing educational content directly applicable in school tutoring. Elementary math tutoring has always been recognized as a demanding process, which is expected to set up the foundations of abstract thinking and to promote interest in exact sciences since childhood. It is important, thus, to enhance and diversify traditional math lessons so that students become more engaged through e-learning and interactive activities. In our experimental implementation of the collaborative e-learning environment model, we provide digitally enhanced educational materials for elementary math tutoring that are suitable for class work, groupwork and self-studies [10]. We implement the augmented interactive desktop as a standard tabletop with digitally enhanced picture cards and an RFID-CLUSPI reader. A single student could use the cards to answer automatically generated elementary math problems by pointing and clicking with the RFIDCLUSPI reader on the cards. In a group of two or more students, participants could take turns to formulate and to solve problems. In a class, the teacher could formulate problems, shown on a large screen and students could register their answers simultaneously. Digitally enhanced picture cards could also be used for building educational games, such as the elementary math domino shown in Fig.9. Figure 9. Elementary math domino. In all cases, the system obtains feedback about the current user by sensing his RFID tag. This functionality is highly desirable even for individual use, since such transparent user authentication gives instant access to the user profile and his learning history. Changeable data related to the current user could be stored locally by the application, in the user database or on the RFID tag. In the first two cases, when user changes computers, e.g. a school computer and a home computer, the stored learning history has to be explicitly communicated over the network or transferred by a memory stick, etc. When the RFID tag is used, however, the user history always stays with him, embedded in his RFID, so much simpler infrastructure becomes possible. 5. Conclusions and future work In this work we have reported a new approach for building digitally enhanced collaborative elearning environments incorporating transparent user recognition and authentication through RFID tags and biometric data. The approach has been applied to the implemented augmented interactive desktop and will be employed in adaptive elementary math tutoring experiments in collaboration with teachers at local elementary schools. References [1] Kanev, K., Turk, D., Orr, T., Brine, J., A Dynamic Group Environment for Collaborative Language Learning. In Proceedings of the 15th International Conference on Computers in Education (ICCE 2007), Hiroshima, Japan, November 5-9, 2007, pp.151-158. [2] Kanev, K., Orr, T., Enhancing Paper Documents with Direct Access to Multimedia for an Intelligent Support of Reading, In Proceedings of the IEEE Conference on the Convergence of Technology and Professional Communication (IPCC 2006), Saratoga Springs, New York, USA, 23-25 Oct. 2006. [3] Jain, A., Hong, L., Pankanti, S., Biometric Identification, Communications of the ACM, Vol. 43, No. 2, pp. 91-98, February, 2003. [4] Jones, L. A., Anton, A. I., Earp, J. B., Towards Understanding User Perceptions of Authentication Technologies, In Proceedings of the 2007 ACM workshop on Privacy in electronic society WPES’07, October 29, 2007, Alexandria, Virginia, USA, pp.91-98. [5] Want, R., The magic of RFID, Queue, Vol.2, No. 7, pp. 40-48, October, 2004. [6] Lahiri, S., RFID Source Book, IBM Press, 2006. [7] Graafstra, A., Build It: An RFID Keyboard, PC Magazine, August, 2006. [8] Kanev, K., Kimura, S., Direct Point-and-Click Functionality for Printed Materials, The Journal of Three Dimensional Images, Vol. 20, No. 2, pp. 51-59, 2006. [9] Kanev, K., Kimura, S., Kobayashi, N., Yamauchi, K., Employment of Physical Objects as Interactive Interface Components. In Proceedings of CLIHC2007 held in conjunction with the 11th IFIP TC 13 Conference on HumanComputer Interactions (INTERACT’2007), Rio de Janeiro, Brazil, September 10-14, 2007. [10] Kanev, K., Kamiya, N., Mirenkov, N., Digitally Enhanced Printed Handouts for Elementary Math Tutoring, In Proceedings of the Fourth International Conference on New Exploratory Technologies (NEXT 2007), Seoul, Korea, October 25-27, 2007, pp.227-230.