Privacy risk quantification in education data using Markov model

Abstract

With Big Data revolution, the education sector is being reshaped. The current data‐driven education system provides many opportunities to utilize the enormous amount of collected data about students' activities and performance for personalized education, adapting teaching methods, and decision making. On the other hand, such benefits come at a cost to privacy. For example, the identification of a student's poor performance across multiple courses. While several works have been conducted on quantifying the re‐identification risks of individuals in released datasets, they assume an adversary's prior knowledge about target individuals. Most of them do not utilize all the available information in the datasets. For example, event‐level information that associates multiple records to the same individual and correlation between attributes. In this work, we propose a method using a Markov Model (MM) to quantify re‐identification risks using all available information in the data under a more realistic threat model that assumes different levels of an adversary's knowledge about the target individual, ranging from any one of the attributes to all given attributes. Moreover, we propose a workflow for efficiently calculating MM risk which is highly scalable to large number of attributes. Experimental results from real education datasets show the efficacy of our model for re‐identification risk. What is already known about this topic? There has been a number of works/research conducted on privacy risk quantification in datasets and in the Web. Majority of them have strong assumption about adversary's prior knowledge of target individual(s). Most of them do not utilize all the available information in the datasets, eg, event‐level or duplicate records and correlation between attributes. What this paper adds? This paper proposes a new re‐identification risk quantification model using Markov models. Our model addresses the shortcomings of existing works, eg, strong assumption about adversary's knowledge, unexplainable model, and utilizing available information in the datasets. Specifically, our proposed model not only focuses on the uniqueness of data points in the datasets (as most of the other existing methods), but also takes into account uniformity and correlation characteristics of these data points. Re‐identification risk quantification is computationally expensive and is not scalable to large datasets with increasing number of attributes. This paper introduces a workflow for data custodians to use to efficiently evaluate the worst‐case re‐identification risk in their datasets before releasing. It presents extensive experimental evaluation results of the proposed model for quantifying re‐identification risks on several real education datasets. Implications for practice and/or policy? Empirical results on real education datasets validate the significance and efficacy of the proposed model for re‐identification risk quantification compared to existing approaches. Our model can be used by the data custodians as...

Implementing next generation privacy and ethics research in education technology

“…○ On-the-fly estimation of privacy risk reduction with the inclusion of a ready reckoner, which calculates the change to privacy risk with the implementation of additional deidentification steps, and ○ the expansion of privacy risk measurement strategies to include measures that look across time series or other sequentially linked data for measures of uniformity and correlation of identifiable patterns [Vatsalan (2022) in preparation for this issue]. • Implementation of new forms of provable privacy risk reduction mechanisms for commonly used queries such as "top/bottom k" (works in preparation, available on request).…”

“…In this paper, we described a framework to implement privacy‐preserving LA, which pragmatically leverages existing technology and knowhow, and includes new achievements which should further help to make ethical, privacy‐preserving LA a practical reality: Augmentations to the PRE now allow: ○On‐the‐fly estimation of privacy risk reduction with the inclusion of a ready reckoner, which calculates the change to privacy risk with the implementation of additional de‐identification steps, and ○the expansion of privacy risk measurement strategies to include measures that look across time series or other sequentially linked data for measures of uniformity and correlation of identifiable patterns [Vatsalan (2022) in preparation for this issue]. Implementation of new forms of provable privacy risk reduction mechanisms for commonly used queries such as “top/bottom k” (works in preparation, available on request). Development of an LA library, for use with our privacy risk reduction treatments, which will help the LA community adopt privacy‐preserving analytics (“Preserving Both Privacy and Utility in Learning Analytics”, Chen Zhan et al in preparation for this issue). This continues to be expanded. Learning Analytics studies which showed interesting results in a number of areas (Joksimovic et al., 2020 for example).…”

“…The PRE calculates re‐identification risk from several statistical probabilities, looking at a single attribute (column) in a dataset and combinations of attributes. Its original re‐identification risk measure then uses information theory to calculate the uniqueness of a specific attribute or combination of attributes in the dataset (factoring in the overall size of the dataset); and a more recent, additional, measure using probabilistic modelling (using Markov chains) to calculate re‐identification risk based on uniformity and patterns of correlation in time series data [Vatsalan (2022) in preparation for this issue].…”

Is it time we get real? A systematic review of the potential of data-driven technologies to address teachers' implicit biases

“…Although sources of bias that relate to the algorithmic and/or data-related biases are of significant concern, they are not within the scope of this research review. There is a growing body of research aiming to address some of these issues covered more from a technical point of view (e.g., Deho et al, 2022 ; Vatsalan et al, 2022 ). For instance, Kizilcec and Lee (forthcoming) proposed critical fairness criteria of independence, separation and sufficiency for potentially mitigating bias in the predictions of AI applications in Education.…”

Technological frameworks on ethical and trustworthy learning analytics

“…As indicated by Marshall et al (2022), in addition to the infrastructure, a crucial aspect of developing trustworthy LA is centred around the algorithms and methods that would allow us to measure and mitigate potential risks. In that sense, Vatsalan et al (2022) propose a new re-identification risk measure based on a Markov Model that does not require strong assumptions about the adversary's knowledge of the datasets. The method not only focuses on the uniqueness of data points but also considers the uniformity and correlation characteristics of these data points.…”