This paper investigates facilitating remote collection of a patient’s data in distributed system while protecting the security of the data, preserving the privacy of the patient’s ID, and preventing inference attack. The paper presents a novel framework called SPID stand for a Secure, ID Privacy, and Inference Threat Prevention Mechanisms for Distributed Systems. In designing this framework, we make the following novel contributions. The SPID presents a novel architecture that supports the use of a distributed set of servers owned by different service providers. The SPID allows the patient to access these servers using certificates generated by the patient. The SPID allows the patient to select one server to be the home server, and select a number of servers to be the foreign servers. The patient uses the foreign servers to upload data. The home server is responsible for collecting the patient’s data from the foreign servers and sending them to the healthcare provider. The SPID proposes a method for efficient verification of each request from the patient without searching in the server’s database for the verification key. This is done by using some of the Elliptic Curves Cryptography (ECC) properties. The SPID has been analyzed using a bench-marking tool and evaluated using queuing theory. The evaluation results indicate an efficient performance when the number of servers increases. We uses Shannon entropy method to measure the likelihood of the inference attack.

Deciding upon instructor intervention based on learners' comments that need an urgent response in MOOC environments is a known challenge. The best solutions proposed used automatic machine learning (ML) models to predict the urgency. These are 'black-box'-es, with results opaque to humans. EXplainable artificial intelligence (XAI) is aiming to understand these, to enhance trust in artificial intelligence (AI)-based decision-making. We propose to apply XAI techniques to interpret a MOOC intervention model, by analysing learner comments. We show how pairing a good predictor with XAI results and especially colour-coded visualisation could be used to support instructors making decisions on urgent intervention.

Automatically identifying the learner gender, which serves as this paper's focus, can provide valuable information to personalised learners' experiences in MOOCs. However, extracting the gender from learner-generated data (discussion forum) is a challenging task, which is understudied in literature. Using syntactic features is still the state-of-the-art for gender identification in social media. Instead we propose here a novel approach based on Recursive Neural Networks (RecNN), to learn advanced syntactic knowledge extracted from learners' comments, as an NLP-based predictor for their gender identity. We propose a bi-directional composition function, added to NLP state-of-the-art candidate RecNN models. We evaluate different combinations of semantic level encoding and syntactic level encoding functions, exploring their performances, with respect to the task of learner gender profiling in MOOCs.

Massive Open Online Courses (MOOCs) have become universal learning resources, and the COVID-19 pandemic is rendering these platforms even more necessary. These platforms also bring incredible diversity of learners in terms of their traits. A research area called Author Profiling (AP in general; here, Learner Profiling (LP)), is to identify such traits about learners, which is vital in MOOCs for, e.g., preventing plagiarism, or eligibility for course certification. Identifying a learner’s trait in a MOOC is notoriously hard to do from textual content alone. We argue that to predict a learner’s academic level, we need to also be using other features stemming from MOOC platforms, such as derived from learners’ actions on the platform. In this study, we specifically examine time stamps, quizzes, and discussions. Our novel approach for the task achieves a high accuracy (90% in average) even with a simple shallow classifier, irrespective of data size, outperforming the state of the art.

Massive Open Online Courses (MOOCs) have become universal learning resources, and the COVID-19 pandemic is rendering these platforms even more necessary. In this paper, we seek to improve Learner Profiling (LP), i.e. estimating the demographic characteristics of learners in MOOC platforms. We have focused on examining models which show promise elsewhere, but were never examined in the LP area (deep learning models) based on effective textual representations. As LP characteristics, we predict here the employment status of learners. We compare sequential and parallel ensemble deep learning architectures based on Convolutional Neural Networks and Recurrent Neural Networks, obtaining an average high accuracy of 96.3% for our best method. Next, we predict the gender of learners based on syntactic knowledge from the text. We compare different tree-structured Long-Short-Term Memory models (as state-of-the-art candidates) and provide our novel version of a Bi-directional composition function for existing architectures. In addition, we evaluate 18 different combinations of word-level encoding and sentence-level encoding functions. Based on these results, we show that our Bi-directional model outperforms all other models and the highest accuracy result among our models is the one based on the combination of FeedForward Neural Network and the Stack-augmented Parser-Interpreter Neural Network (82.60% prediction accuracy). We argue that our prediction models recommended for both demographics characteristics examined in this study can achieve high accuracy. This is additionally also the first time a sound methodological approach toward improving accuracy for learner demographics classification on MOOCs was proposed.

Identifying users’ demographic characteristics is called Author Profiling task (AP), which is a useful task in providing a robust automatic prediction for different social user aspects, and subsequently supporting decision making on massive information systems. For example, in MOOCs, it used to provide personalised recommendation systems for learners. In this paper, we explore intelligent techniques and strategies for solving the task, and mainly we focus on predicting the employment status of users on a MOOC platform. For this, we compare sequential with parallel ensemble deep learning (DL) architectures. Importantly, we show that our prediction model can achieve high accuracy even though not many stylistic text features that are usually used for the AP task are employed (only tokens of words are used). To address our highly unbalanced data, we compare widely used oversampling method with a generative paraphrasing method. We obtained an average of 96.4% high accuracy for our best method, involving sequential DL with paraphrasing overall, as well as per-individual class (employment statuses of users).

Mobile patient monitoring systems monitor and treat chronic diseases by collecting health data from wearable sensors through mobile devices carried out by patients. In the future, these systems may be hosted by a third-party service provider. This would open a number of security and ID privacy issues. One of these issues is the inference attack. This attack allows a single service provider from inferring the patient’s identity by collecting a number of contextual information about the patient such as the pattern of interaction with the service provider. Thus a security and ID privacy mechanisms must be deployed. In this paper, we propose a framework called Secure and Privacy-Preserving Data Collection (SPDC) that allows the patient to encrypt the data and then upload the encrypted data on different service providers rather than one while allowing an anonymous linkage for the patient’s data which are scattered across different service providers. In this framework, each patient is allowed to select the service providers involved in the data collection, assigns one as the home while the others consider foreign. The patient uses the foreign to upload data while the home is responsible for anonymously collecting the patient’s data from multiple foreign service providers and deliver them to the healthcare provider. This framework also shows a novel mechanism to conduct anonymous authentication across different distributed service provides. The framework has been analyzed against the specified design requirements and security threats.

Author Profiling (AP), which aims to predict an author's demographics characteristics automatically by using texts written by the author, is an important mechanism for many applications, as well as highly challenging. In this research, we analyse various previous machine learning models for AP, with respect to their potential for our research problem. Based on this, we propose a Deep Learning Architecture to predict the demographics characteristics of the learners in MOOCs, incorporating multi-feature representations and ensemble learning methods. Specifically, we employ a novel pipeline, combining the most successful deep learning classifiers, Convolution Neural Networks, Recurrent Neural Networks and Recursive Neural Networks, to learn from a text. Moreover, beside the state-of-the-art training involving character and word-level input, we additionally propose phrase-level input. With this approach, we aim at deepening our understanding of the writing style of learners, and thus, predict the author profile with high accuracy. In this paper, we propose the model and architecture, and report on initial tests of our model on a large dataset from the FutureLearn platform, to predict the demographics characteristics of the learners.

Third-party based mobile health monitoring systems are vulnerable to threats not only imposed by outsiders but also authorized insiders, e.g. employees of the third-party service provider. This paper examines issues in this context and proposes a novel framework, called a secure and ID privacy preserving framework (SPDC), to facilitate secure and ID privacy-preserving data collections from remote patients. The framework has a number of properties: (i) it supports three different modes of data collections and different treatment of data with two levels of urgency, (ii) it protects the confidentiality  and authenticity of data being collected, (iii) it preserves patients' ID privacy against external entities as well as entities that are involved in facilitating the data collections, (iv) it uses distributed load-sharing so that no single entity is wholly responsible for the task of data collections. The ideas used in achieving these properties include the use of  structured use of multiple data collection servers  and multi-level hierarchical pseudonyms to hide patients' IDs as well as their communication patterns, and the separation of duties and pseudonym and blind-token based authentication methods so that patients' can be identified and authenticated without revealing their real IDs. Security analysis against design requirements and anonymity evaluation using entropy method are carried out to demonstrate that the framework can resists attacks on data security and protects the patients' IDs from being revealed to unauthorized entities.

 Abstract — The more an educational system knows about a learner, the more personalised interaction it can provide, which leads to better learning. However, asking a learner directly is potentially disruptive, and often ignored by learners. Especially in the booming realm of MOOC Massive Online Learning platforms, only a very low percentage of users disclose demographic information about themselves. Thus, in this paper, we aim to predict learners’ demographic characteristics, by proposing an approach using linguistically motivated Deep Learning Architectures for Learner Profiling, particularly targeting gender prediction on a FutureLearn MOOC platform. Additionally, we tackle here the difficult problem of predicting the gender of learners based on their comments only – which are often available across MOOCs. The most common current approaches to text classification use the Long Short-Term Memory (LSTM) model, considering sentences as sequences. However, human language also has structures. In this research, rather than considering sentences as plain sequences, we hypothesise that higher semantic - and syntactic level sentence processing based on linguistics will render a richer representation. We thus evaluate, the traditional LSTM versus other bleeding edge models, which take into account syntactic structure, such as tree-structured LSTM, Stack-augmented Parser-Interpreter Neural Network (SPINN) and the Structure-Aware Tag Augmented model (SATA). Additionally, we explore using different word-level encoding functions. We have implemented these methods on Our MOOC dataset, which is the most performant one comparing with a public dataset on sentiment analysis that is further used as a cross-examining for the models' results.