Traditional physical biometrics—such as fingerprints, facial recognition, and iris scans—have long been utilized for user identification in areas like border control, military operations, law enforcement, and public safety. However, the rise of smartphone technology has introduced new avenues for security research. One emerging area is behavioral biometrics, particularly the use of touchscreen interaction data as a more accessible and user-friendly identification method. In this context, our study focuses on a novel form of touchscreen input: capacitive swipe gestures for user identification. We compiled a comprehensive dataset of capacitive swipe gestures collected over multiple sessions from 30 participants. To evaluate this modality, we conducted thorough experiments using established machine learning algorithms, including Support Vector Machine, Random Forest, and XGBoost. Additionally, we developed a new preprocessing algorithm tailored for capacitive swipe data. Our findings reveal that this algorithm significantly enhances identification performance compared to existing methods. Overall, our results highlight the strong potential of capacitive swipe gestures as a viable biometric modality for user identification.

In this paper, we focus on improving the age estimation accuracy on smartphones. Estimating a smartphone user’s age has several applications such as protecting our children online by filtering age-inappropriate contents, providing a customized e-commerce experience, etc. However, accuracy of the the state-of-the-art age estimation techniques that use touch behavior on smartphones is still limited because of the lack of sufficient amount of training data. We perform rigorous experiments using zoom gestures on smartphones and demonstrate that increasing the amount of training data can significantly improve the age estimation accuracy. Based on the findings in this study, we recommend creating a large touch dynamics-based age estimation data set so that more accurate age estimation models can be built and in turn, can be used more confidently.

Authorship attribution identifies the true author of an unknown document. Authorship attribution plays a crucial role in plagiarism detection and blackmailer identification, however, the existing studies on authorship attribution in Bengali are limited. In this paper, we propose an instance-based deep authorship attribution model, called DAAB, to identify authors in Bengali. Our DAAB model fuses features from convolutional neural networks and another set of features from an artificial neural network to learn the stylometry of an author for authorship attribution. Extensive experiments with three real benchmark datasets such as Bengali-Quora and two online Bengali Corpus demonstrate the superiority of our authorship attribution model.

Adult content on the Internet may be accessed by children with only a few keystrokes. While separate child-safe accounts may be established, a better approach could be incorporating automatic age estimation capability into the browser. We envision a safer browsing experience by implementing child-safe browsers combined with Internet content rating similar to the film industry. Before such a browser is created it was necessary to test the age estimation module to see whether acceptable error rates are possible. We created an Android application for collecting biometric touch data, specifically tapping data. We arranged with an elementary school, a middle school, a high school, and a university and collected samples from 262 user sessions (ages 5 to 61). From the tapping data, feature vectors were constructed, which were used to train and test 14 regressors and classifiers. Results for regression show the best mean absolute errors of 3.451 and 3.027 years, respectively, for phones and tablets. Results for classification show the best accuracies of 73.63% and 82.28%, respectively, for phones and tablets. These results demonstrate that age estimation, and hence, a child-safe browser, is feasible, and is a worthwhile objective.

Traditional keyboards remain the input device of choice for typing-heavy environments. When attached to sensitive data, security is a major concern. To continuously authenticate users in these environments, use of keystroke dynamics can be a preferred choice. An integral part of user enrollment in a keystroke based continuous authentication system is the writing instruction (prompt) given to the users, to use as a basis for their improvised writing. There are many prompts possible, and they directly impact the performance of authentication systems. Hence, prompts should be designed carefully, and with purpose. In this paper, we bridge the gap between cognitive psychology and computer science and attempt to influence the mental state of the users to acquire a better authentication performance. We compare two kinds of writing prompts, creative and factual, for generating reference samples. In addition, we perform two robustness tests: robustness to dissimilar writing style (e.g., creative reference and factual test) and robustness to surface (e.g., hard surface reference and soft surface test). We collect data from thirty participants in four weekly sessions. We experiment with three features: key interval, key press, and key hold latencies. We use Relative (R) measure to generate the match score between the reference and test samples. Results show that creative writing consistently performs better than the factual one. Both writing prompts perform well with dissimilar style in testing, i.e., continuous authentication is found robust to writing style. Also, we find that the surface (hard or soft) used in testing need not match that used for the reference, thus continuous authentication is also surface robust.

The cyber-behavioral biometric modalities such as keystroke dynamics, mouse dynamics, and touch screen dynamics have come under attacks of different forms in recent days. To address these attacks and other security issues, we present a novel concept of using smartwatch sensor data to continuously verify users in cyberspace and show its potential to be a new standalone cyber-behavioral biometric modality. For our experiments, smartwatch gyroscope and accelerometer data collected from 49 subjects while typing in desktop computer have been considered. We implemented six pattern matching classifiers to compare each verification attempt against the user profile. Experimental results comprising of 282,240 classification attempts show significantly high True Positive (TP) rates and extremely low False Positive (FP) rates with the highest achieved TP rate of 87.2% and lowest FP rate of 0.2%. With this level of accuracy and natural resiliency to attacks comes with physical biometric property as such in hand movement, we opine that smartwatch movement dynamics, besides being a new biometric trait, can be a solution to the security loopholes in existing cyber-behavioral biometric modalities for continuous verification.

Keystroke dynamics has been used as a form of one-time user authentication and continuous verification especially when it comes to securing the cyberspace. In this paper, we present the idea of using keystroke dynamics as a form of second layer authentication in web applications. We showed that this method can authenticate a user with high accuracy and can be used as an alternate to CAPTCHA tests, security questions and image selections that are being used today. We have developed a working web-based platform in a browser environment that enforces the proposed second-layer security. We performed penetration test experiments by launching a total of 598,500 impostor and genuine authentication attempts and found the Equal Error Rate (EER) as 10.5%.

A smartphone-based fall detection system has two major advantages over a traditional fall detection system that comes as a separate device: (1) the phone can automatically send messages to or call the emergency contact person when a fall is detected and (2) a user does not need to carry an extra device. This paper presents a novel two-step fall detection method which uses data extracted from smartphone sensors to detect falls. A fall can happen in many ways. A person can fall while he/she is walking, jogging, sitting, or even sleeping. Patterns of all falls are not the same. It is important to identify the type of falls to precisely distinguish it from non-falls (normal activities). Hence, our method first identifies the correct type of falls by performing multi-class classification. In the second step, this method produces a binary decision based on the multiclass prediction. We collected data from 10 users to evaluate our proposed fall detection method. Each user performed five normal activities-namely, walking, jogging, standing, sitting, lying, and also fell after performing each activity. We performed experiments with five common smartphone sensors: accelerometer, gyroscope, magnetometer, gravity, and linear acceleration. We tested five machine learning classifiers-namely, Support Vector Machine, K-Nearest Neighbor, Decision Tree, Random Forest, and Naive Bayes. Our two-step fall detection method achieved the maximum accuracy of 95.65% and the maximum area under ROC curve (AUC) of 0.93, both with the gyroscope sensor and Support Vector Machine classifier.

We outline a novel method of user authentication for smart mobile devices, such as smartphones or tablets and propose movement pattern based authentication as an alternate to current methods that relies on a pin or drawn-pattern. While the current methods are vulnerable against common attacks (e.g., smudge attacks, shoulder surfing), our method, in contrast, is more resilient against the attacks of these kinds because it utilizes sensory data given off by the device during a preset movement for authentication. In our experiment, we recorded the values given off by four physical observational sensors: (1) accelerometer, (2) linear accelerometer, (3) gyroscope and (4) tilt sensor, which each had three axes, over a set of movements. We experimented with 10 arbitrary movement-patterns and gathered 12 samples of each (net 120 samples) to test with. We developed our own method of authentication, through which we performed 35,650 authentication attempts and found a 20.36% Equal Error Rate.

The traditional architecture of serial fusion based multi-biometric verification systems places an average performing or the worst performing individual verifier in the final stage. Because the final stage gives the verification decision using a single threshold and takes on the most confusing samples which are rejected by all previous stages, an average or the worst performing individual verifier may incur high verification errors in the final stage, which may negatively impact the performance of the whole system. Unfortunately, it is not possible to place a strong individual verifier in the final stage of a traditional architecture because if we place a strong individual verifier in the final stage, we will have to place a weak individual verifier in an earlier stage. Studies show that placing a weak individual verifier in an earlier stage worsens the performance of the whole system by giving more wrong decision earlier. Hence, the challenge is-how can we place the best performing individual verifier in the first stage and at the same time not place an average or the worst performing individual verifier in the final stage? In this paper, we address this challenge. We have come up with a very simple but effective solution. We have proposed a modification to the traditional architecture of serial fusion based multi-biometric verification systems. With rigorous experiments on the NIST multi-modal dataset and using three serial fusion based multi-biometric verification schemes, we demonstrated that our proposed architecture significantly improves the performance of serial fusion based multi-biometric verification systems.

In this paper, we study the influence from the sentiment of regular tweets on retweeting. We propose a method to calculate the sentiment score for each tweet and each Twitter user. This method enables us to place the tweets and retweets into the same time period to explore the sentiment factor. We adopt the correlation coefficient between the sentiment scores of regular tweets and those of retweets to measure the influence. We categorize the Twitter users in three different ways to investigate three factors, which are the number of followers, betweenness centrality and the types of accounts. Community detection and machine learning are integrated into our approach. We find that the difference for correlation coefficients exists between different levels of the number of followers, and different types of users. Our method sheds a light on better predicting the dynamics of tweets diffusion by including the sentiment factor into the prediction model.

Smartphones, while providing users ease of access to sensitive information on the go, also present severe security risks if an attacker is able to gain access to them. To strengthen the user authentication and identification in a smartphone, we develop a biometric authentication and identification system which uses the capacitive touchscreen that is featured in all current smartphones. Our methodology focuses on using the touchscreen as a sensor to capture the image of a user's ear, thumb or four fingers. We extract the capacitive raw data from the touched body part to obtain a capacitive image, and then use it to capture geometric features (e.g., length and width of a finger) and principal components. After that, we experiment with Support Vector Machine (SVM) and Random Forest (RF) classifiers to verify and also identify each user. We achieved the maximum authentication accuracy of 98.84% by four fingers with SVM, and maxinum identification accuracy of 97.61% by four fingers with RF.