Search

Computer networks form much of the infrastructure supporting day-to-day life in this digital age. Computer networks, however, are prone to attack and therefore require intrusion detection systems. Intrusion detection systems provide a mechanism to detect network attacks at an early stage and generate alerts. These systems, however, are far from a panacea. Rather, they tend to overwhelm their operators with alerts, which in more than 90% of cases can be false positives. As such, the problem of false positives in intrusion detection systems is a costly issue. This paper presents research to design a hierarchical network intrusion detector, using deep learning, which protects against raising vast numbers of false positives through the design and implementation of a hierarchical NIDS. This paper presents a valuable advancement in performance by reducing the occurrence of false alarms by 87.52%. The research contained in this paper presents three contributions to knowledge. The first of these is the comparison between hierarchical systems and non-hierarchical systems to understand which would yield fewer false alarms. The second contribution is the formulation of a hierarchical approach, which was able to reduce false alarms by 87.52%. Lastly, the proposed hierarchical model was deployed in a live IoT environment, exposed to genuine threats, and the performance in this environment was analysed.

Abstract Background Office work generally consists of high amounts of sedentary behaviour (SB) which has been associated with negative health consequences. We developed the “WorktivIty” mobile app to help office workers reduce their SB through self-monitoring and feedback on sedentary time, prompts to break sedentary time, and educational facts. The aim of this paper is to report the feasibility of delivering the Worktivity intervention to desk-based office workers in the workplace setting and describe methodological considerations for a future trial. Methods We conducted a three-arm feasibility cluster randomised controlled pilot study over an 8-week period with full time-desk based employees. Clustered randomisation was to one of three groups: Worktivity mobile app (MA; n = 20), Worktivity mobile app plus SSWD (MA+SSWD; n = 20), or Control (C; n = 16). Feasibility was assessed using measures of recruitment and retention, intervention engagement, intervention delivery, completion rates and usable data, adverse events, and acceptability. Results Recruitment of companies to participate in this study was challenging (8% of those contacted), but retention of individual participants within the recruited groups was high (81% C, 90% MA + SSWD, 95% MA). Office workers’ engagement with the app was moderate (on average 59%). Intervention delivery was partially compromised due to diminishing user engagement and technical issues related to educational fact delivery. Sufficient amounts of useable data were collected, however either missing or unusable data were observed with activPAL™, with data loss increasing at each follow up time point. No serious adverse events were identified during the study. The majority of participants agreed that the intervention could be implemented within the workplace setting (65% MA; 72% MA + SSWD) but overall satisfaction with the intervention was modest (58% MA; 39% MA + SSWD). Conclusions The findings suggest that, in principle, it is feasible to implement a mobile app-based intervention in the workplace setting however the Worktivity intervention requires further technical refinements before moving to effectiveness trials. Challenges relating to the initial recruitment of workplaces and maintaining user engagement with the mHealth intervention over time need to be addressed prior to future large-scale implementation. Further research is needed to identify how best to overcome these challenges.

Smartphone-based approaches for Human Activity Recognition have become prevalent in recent years. Despite the amount of research undertaken in the field, issues such as cross-subject variability are still posing an obstacle to the deployment of solutions in large scale, free-living settings. Personalized methods (i.e. aiming to adapt a generic classifier to a specific target user) attempt to solve this problem. The lack of labeled data for training purposes, however, represents a major barrier. This is especially the case when taking into consideration that personalization generally requires labeled data to be user-specific. This paper presents a novel personalization method combining a semi-population based approach with user adaptation. Personalization is achieved through the following. Firstly, the proposed method identifies a subset of users from the available population as best candidates for initializing the classifier to the target user. Subsequently, a semi-population Neural Network classifier is trained using data from this subset of users. The classifier's network weights are then updated using a small amount of labeled data from the target user subsequently implementing personalization. This approach was validated on a large publicly available dataset collected in a free-living scenario. The personalized approach using the proposed method has shown to improve the overall F-score to 74.4% compared to 70.9% when using a generic non-personalized approach. Results obtained, with statistical significance being confirmed on a set of 57 users, indicate that model initialization using the semi-population approach can reduce the amount of labeled data required for personalization. As such, the proposed method for model initialization could facilitate the real-world deployment of systems implementing personalization by reducing the amount of data needed for personalization.

Activity Recognition is an active research topic focused on detecting human actions and behaviours in smart environments. In most cases, the use of data-driven models aim to relate data from sensors to an activity through a model developed by a supervised approach. In this work, we focus on the goal of domain adaptation between smart environments, which has required a novel approach to relate the concepts of domain adaptation using binary sensor and learning from daily imbalanced data. In this work, the sensor activation from a given context is translated to a different one, based on the temporal alignment from human activities. The domain adaptation of binary sensor is accomplished through a three step procedure: i) clustering of sensor activation, ii) activity based alignment of sensor data between the two environments, iii) an ensemble of classifiers used to mine a mapping function, translating sensor data between the two environments. The proposed method was evaluated over a publicly available dataset, and obtained preliminary results which were encouraging with an F1-Score of 87%.

Research interests spanning numerous domains increasingly rely upon computational systems which can store and process a large volume of variable data that is stored at high velocity-representing a big data problem. This is particularly notable within the domain of ubiquitous and pervasive computing. This domain increasingly relies on storage and retrieval of sensor data to enable outcomes such as predictive analytics and activity recognition. Several current big data platforms exist; however, they have a range of deficiencies including lack of generic interoperability with agnostic sensors and an absence of features supporting academic research. Due to these deficiencies a custom, research oriented, high performance, and big data platform was devised and implemented. This platform is called SensorCentral and is presented within this paper. SensorCentral provides a framework which enables interoperability with a large range of agnostic sensor devices whilst simultaneously providing features which support research. Research supporting features include; facility to define experiments, ability to annotate experimental instances via purpose-built mobile applications, integrated machine learning functionality, facility to export data sets, rule-based classification and an extensible platform. The flagship implementation of this platform has been in operation for over 28 months within a University research group and has been successfully integrated with a range of sensors from a variety of manufacturers. This implementation currently stores over 850 million records and has been central to several research and industrial projects. Future work will integrate this platform into the open data initiative enabling collaboration with the international community of researchers.

Abstract Background High levels of sedentary behaviour (SB) are associated with negative health consequences. Technology enhanced solutions such as mobile applications, activity monitors, prompting software, texts, emails and websites are being harnessed to reduce SB. The aim of this paper is to evaluate the effectiveness of such technology enhanced interventions aimed at reducing SB in healthy adults and to examine the behaviour change techniques (BCTs) used. Methods Five electronic databases were searched to identify randomised-controlled trials (RCTs), published up to June 2016. Interventions using computer, mobile or wearable technologies to facilitate a reduction in SB, using a measure of sedentary time as an outcome, were eligible for inclusion. Risk of bias was assessed using the Cochrane Collaboration’s tool and interventions were coded using the BCT Taxonomy (v1). Results Meta-analysis of 15/17 RCTs suggested that computer, mobile and wearable technology tools resulted in a mean reduction of −41.28 min per day (min/day) of sitting time (95% CI -60.99, −21.58, I2 = 77%, n = 1402), in favour of the intervention group at end point follow-up. The pooled effects showed mean reductions at short (≤ 3 months), medium (>3 to 6 months), and long-term follow-up (>6 months) of −42.42 min/day, −37.23 min/day and −1.65 min/day, respectively. Overall, 16/17 studies were deemed as having a high or unclear risk of bias, and 1/17 was judged to be at a low risk of bias. A total of 46 BCTs (14 unique) were coded for the computer, mobile and wearable components of the interventions. The most frequently coded were “prompts and cues”, “self-monitoring of behaviour”, “social support (unspecified)” and “goal setting (behaviour)”. Conclusion Interventions using computer, mobile and wearable technologies can be effective in reducing SB. Effectiveness appeared most prominent in the short-term and lessened over time. A range of BCTs have been implemented in these interventions. Future studies need to improve reporting of BCTs within interventions and address the methodological flaws identified within the review through the use of more rigorously controlled study designs with longer-term follow-ups, objective measures of SB and the incorporation of strategies to reduce attrition. Trial registration The review protocol was registered with PROSPERO: CRD42016038187

Activity recognition has received considerable attention in many research fields, such as industrial and healthcare fields. However, many researches about activity recognition have focused on static activities and dynamic activities in current literature, while, the transitional activities, such as stand-to-sit and sit-to-stand, are more difficult to recognize than both of them. Consider that it may be important in real applications. Thus, a novel framework is proposed in this paper to recognize static activities, dynamic activities, and transitional activities by utilizing stacked denoising autoencoders (SDAE), which is able to extract features automatically as a deep learning model rather than utilize manual features extracted by conventional machine learning methods. Moreover, the resampling technique (random oversampling) is used to improve problem of unbalanced samples due to relatively short duration characteristic of transitional activity. The experiment protocol is designed to collect twelve daily activities (three types) by using wearable sensors from 10 adults in smart lab of Ulster University, the experiment results show the significant performance on transitional activity recognition and achieve the overall accuracy of 94.88% on three types of activities. The results obtained by comparing with other methods and performances on other three public datasets verify the feasibility and priority of our framework. This paper also explores the effect of multiple sensors (accelerometer and gyroscope) to determine the optimal combination for activity recognition.