Fudong Li

Unobtrusive Gait Recognition Using Smartwatches

Gait recognition is a technique that identifies or verifies people based upon their walking patterns. Smartwatches, which contain an accelerometer and gyroscope have recently been used to implement gait-based biometrics. However, this prior work relied upon data from single sessions for both training and testing, which is not realistic and can lead to overly optimistic performance results. This paper aims to remedy some of these problems by training and evaluating a smartwatch-based biometric system on data obtained from different days. Also, it proposes an advanced feature selection approach to identify optimal features for each user. Two experiments are presented under three different scenarios: Same-Day, Mixed-Day, and Cross-Day. Competitive results were achieved (best EERs of 0.13% and 3.12% by using the Same day data for accelerometer and gyroscope respectively and 0.69% and 7.97% for the same sensors under the Cross-Day evaluation. The results show that the technology is sufficiently capable and the signals captured sufficiently discriminative to be useful in performing gait recognition.

Insider Misuse Identification using Transparent Biometrics

Insider misuse is a key threat to organizations. Recent research has focused upon the information itself – either through its protection or approaches to detect the leakage. This paper seeks a different approach through the application of transparent biometrics to provide a robust approach to the identification of the individuals who are misusing systems and information. Transparent biometrics are a suite of modalities, typically behavioral-based that can capture biometric signals covertly or non-intrusively – so the user is unaware of their capture. Transparent biometrics are utilized in two phases a) to imprint digital objects with biometric-signatures of the user who last interacted with the object and b) uniquely applied to network traffic in order to identify users traffic (independent of the Internet Protocol address) so that users rather than machine (IP) traffic can be more usefully analyzed by analysts. Results from two experimental studies are presented and illustrate how reliably transparent biometrics are in providing this link-ability of information to identity.

Evidence Identification in Heterogeneous Data Using Clustering

Digital forensics faces several challenges in examining and analyzing data due to an increasing range of technologies at peoples disposal. The investigators find themselves having to process and analyze many systems manually (e.g. PC, laptop, Smartphone) in a single case. Unfortunately, current tools such as FTK and Encase have a limited ability to achieve the automation in finding evidence. As a result, a heavy burden is placed on the investigator to both find and analyze evidential artifacts in a heterogenous environment. This paper proposed a clustering approach based on Fuzzy C-Means (FCM) and K-means algorithms to identify the evidential files and isolate the non-related files based on their metadata. A series of experiments using heterogenous real-life forensic cases are conducted to evaluate the approach. Within each case, various types of metadata categories were created based on file systems and applications. The results showed that the clustering based on file systems gave the best results of grouping the evidential artifacts within only five clusters. The proportion across the five clusters was 100% using small configurations of both FCM and K-means with less than 16% of the non-evidential artifacts across all cases -- representing a reduction in having to analyze 84% of the benign files. In terms of the applications, the proportion of evidence was more than 97%, but the proportion of benign files was also relatively high based upon small configurations. However, with a large configuration, the proportion of benign files became very low less than 10%. Successfully prioritizing large proportions of evidence and reducing the volume of benign files to be analyzed, reduces the time taken and cognitive load upon the investigator.

Identifying and predicting the factors affecting end-users’ risk-taking behavior

Purpose n n n n nThe end-user has frequently been identified as the weakest link; however, motivated by the fact that different users react differently to the same stimuli, identifying the reasons behind variations in security behavior and why certain users could be “at risk” more than others is a step toward protecting and defending users against security attacks. This paper aims to explore the effect of personality trait variations (through the Big Five Inventory [BFI]) on users’ risk level of their intended security behaviors. In addition, age, gender, service usage and information technology (IT) proficiency are analyzed to identify what role and impact they have on behavior. n n n n nDesign/methodology/approach n n n n nThe authors developed a quantitative-oriented survey that was implemented online. The bi-variate Pearson two-tailed correlation was used to analyze survey responses. n n n n nFindings n n n n nThe results obtained by analyzing 538 survey responses suggest that personality traits do play a significant role in affecting users’ security behavior risk levels. Furthermore, the results suggest that BFI score of a trait has a significant effect as users’ online personality is linked to their offline personality, especially in the conscientiousness personality trait. Additionally, this effect was stronger when personality was correlated with the factors of IT proficiency, gender, age and online activity. n n n n nOriginality/value n n n n nThe contributions of this paper are two-fold. First, with the aid of a large population sample, end-users’ security practice is assessed from multiple domains, and relationships were found between end-users’ risk-taking behavior and nine user-centric factors. Second, based upon these findings, the predictive ability for these user-centric factors were evaluated to determine the level of risk a user is subject to from an individual behavior perspective. Of 28 behaviors, 11 were found to have a 60 per cent or greater predictive ability, with the highest classification of 92 per cent for several behaviors. This provides a basis for organizations to use behavioral intent alongside personality traits and demographics to understand and, therefore, manage the human aspects of risk.

Facial-Forensic Analysis Tool

Abstract Facial recognition has played an essential role in digital forensics due to the widespread use of digital technology such as CCTV, mobile phones, and digital cameras. Therefore, the growing volume of multimedia files (photos and videos), in particular, are a valuable source of evidence and the ability to identify culprits’ is invaluable. Despite significant efforts that have been given to this area, facial recognition suffers from several drawbacks in achieving recognition. These reasons are caused by photos conditions issues such as bad illumination, facial orientation, facial expression, photo quality, accessories (e.g., hat, glasses), and aging. The Facial-Forensic Analysis Tool (F-FAT) provides a technique that aids the forensic investigation in terms of the automatic facial recognition. It is a holistic system that is developed to collect, exam, and analyse multimedia evidence (photos and videos) using a multi-algorithmic fusion approach to overcome the weaknesses in individual algorithms and achieve a better accuracy for identification. The proposed approach also helps to reduce the cognitive load placed upon the investigator by providing a variety of forensic analyses such as, geo-location, facial modification, and social networks, to enable quicker answers to queries. This tool has also been designed based upon a case management concept that helps to manage the overall system, provide robust authentication, authorization and chain of custody.

Transparent authentication: Utilising heart rate for user authentication

There has been exponential growth in the use of wearable technologies in the last decade with smart watches having a large share of the market. Smart watches were primarily used for health and fitness purposes but recent years have seen a rise in their deployment in other areas. Recent smart watches are fitted with sensors with enhanced functionality and capabilities. For example, some function as standalone device with the ability to create activity logs and transmit data to a secondary device. The capability has contributed to their increased usage in recent years with researchers focusing on their potential. This paper explores the ability to extract physiological data from smart watch technology to achieve user authentication. The approach is suitable not only because of the capacity for data capture but also easy connectivity with other devices — principally the Smartphone. For the purpose of this study, heart rate data is captured and extracted from 30 subjects continually over an hour. While security is the ultimate goal, usability should also be key consideration. Most bioelectrical signals like heart rate are non-stationary time-dependent signals therefore Discrete Wavelet Transform (DWT) is employed. DWT decomposes the bioelectrical signal into n level sub-bands of detail coefficients and approximation coefficients. Biorthogonal Wavelet (bior 4.4) is applied to extract features from the four levels of detail coefficents. Ten statistical features are extracted from each level of the coffecient sub-band. Classification of each sub-band levels are done using a Feedforward neural Network (FF-NN). The 1st, 2nd, 3rd and 4th levels had an Equal Error Rate (EER) of 17.20%, 18.17%, 20.93% and 21.83% respectively. To improve the EER, fusion of the four level sub-band is applied at the feature level. The proposed fusion showed an improved result over the initial result with an EER of 11.25% As a one-off authentication decision, an 11% EER is not ideal, its use on a continuous basis makes this more than feasible in practice.

Smart watch based body-temperature authentication

The advancement of smart devices has led to a steep rise in wearable devices of which smart watches are increasingly gaining popularity in the wearable technology market. Most smart watches have evolved from their first generation to their present generation with increased functionality and capacity. This has led to smart watches gaining popularity and acceptability within the mainstream digital device usage. The first generation of smart watches were fitted with fewer sensors compared to the present day smart watches. The present day smart watch can be used for various activities much more than its tradition usage for health and fitness. These activities includes accepting and declining calls, reading Short Message Service (SMS), listening to music, navigation etc. while smart watches are still advancing technologically, some can function independently while most can be synchronized with smart phones through Bluetooth or Near-Field Communication (NFC). This brings about their easy communication with smart phones. To access the smart watch applications and information, it will be ideal to authenticate the user. Therefore this paper proposed a novel body temperature authentication system, BT-Authen, to authenticate the user by using the body temperature information extracted via a smart watch for continuous and non-intrusive user authentication. The authentication credentials are compared on the smartphone it is paired with before access is granted. To actualise this, the galvanic skins response (GSR) and skin temperature information are extracted for user authentication. The dataset for the evaluation of the body temperature signals are extracted from 30 subjects over three days. Six features are extracted from each of the two body temperature signals. The classification achieved an EER of 3.4 % using a Neural Network Feedforward (NN-FF) classifier. The performance increased to EER of 0.54% after applying a best performance scoring algorithm.

Leveraging Biometrics for Insider Misuse Identification

Insider misuse has become a real threat to many enterprises in the last decade. A major source of such threats originates from those individuals who have inside knowledge about the organization’s resources. Therefore, preventing or responding to such incidents has become a challenging task. Digital forensics has grown into a de-facto standard in the examination of electronic evidence, which provides a basis for investigating incidents. A key barrier however is often being able to associate an individual to the stolen data—especially when stolen credentials and the Trojan defense are two commonly cited arguments. This paper proposes an approach that can more inextricably link the use of information (e.g. images, documents and emails) to the individual users who use and access them through the use of transparent biometric imprinting. The use of transparent biometrics enables the covert capture of a user’s biometric information—avoiding the potential for forgery. A series of experiments are presented to evaluate the capability of retrieving the biometric information through a variety of file modification attacks. The preliminary feasibility study has shown that it is possible to correlate an individual’s biometric information with a digital object (images) and still be able to recover the biometric signal even with significant file modification. Intl. Journal on Cyber Situational Awareness, Vol. 1, No. 1, 2016