A privacy-preserving distributed filtering framework for NLP artifacts

Abstract

Background Medical data sharing is a big challenge in biomedicine, which often hinders collaborative research. Due to privacy concerns, clinical notes cannot be directly shared. A lot of efforts have been dedicated to de-identifying clinical notes but it is still very challenging to accurately locate and scrub all sensitive elements from notes in an automatic manner. An alternative approach is to remove sentences that might contain sensitive terms related to personal information. Methods A previous study introduced a frequency-based filtering approach that removes sentences containing low frequency bigrams to improve the privacy protection without significantly decreasing the utility. Our work extends this method to consider clinical notes from distributed sources with security and privacy considerations. We developed a novel secure protocol based on private set intersection and secure thresholding to identify uncommon and low-frequency terms, which can be used to guide sentence filtering. Results As the computational cost of our proposed framework mostly depends on the cardinality of the intersection of the sets and the number of data owners, we evaluated the framework in terms of these two factors. Experimental results demonstrate that our proposed method is scalable in various experimental settings. In addition, we evaluated our framework in terms of data utility. This evaluation shows that the proposed method is able to retain enough information for data analysis. Conclusion This work demonstrates the feasibility of using homomorphic encryption to develop a secure and efficient multi-party protocol.

A flexible and stretchable bionic true random number generator

“…During the process of random number generation, evaluating the amount of speckle randomness is highly significant for subsequent randomness extraction. The randomness of raw speckle images under different SLM configurations has been assessed by means of the min-entropy, which served as the conservative measure of randomness [47,48]. In our scenario, the min-entropy (E) of an 8-bit greyscale speckle image can be defined as follows…”

Privacy-Preserving Deep Learning NLP Models for Cancer Registries

“…Since manual de-identification approach is costly and time consuming, different techniques have been proposed to support automatic clinical text de-identification using traditional machine learning [11] and DL models [12]. However, automatically locating and scrubbing all sensitive information from clinical text is still highly challenging -deidentifying unstructured text in pathology reports is more challenging than structured data [13], and de-identification models trained on a specific dataset do not generalize well to other datasets [14]. Existing solutions typically cannot guarantee de-identification up to regulatory standards, especially with scattered PHI across the unstructured text of pathology reports; therefore, there is still need for alternative privacy-preserving methods to protect PHI and directly or indirectly share large corpora of pathology reports from various sources.…”

Information security implications of using NLP in IT outsourcing: a Diffusion of Innovation theory perspective

“…For example, as the industry is striving to automate ITO service delivery, NLP can help to improve the efficiency of collaboration among the client and service provider teams or to identify vulnerabilities in the software specification documentation produced by service provider organisations for their clients (Kang et al 2020;Zhang et al 2019). Similarly, NLP can be used to ensure privacy-preservation in multi-party ITO relationships (Sadat et al 2019). However, NLP comes with its own implications related to information security (Zhang et al 2019).…”

“…However, NLP comes with its own implications related to information security (Zhang et al 2019). The investigation of those implications is swiftly gaining the attention of researchers as the use of NLP for improving ITO experience results in information security vulnerabilities that are not yet fully understood (Alshemali and Kalita 2020;Feyisetan et al 2020;Sadat et al 2019;Winter and Rinderle-Ma 2018).…”

“…Kang et al (2020) andWinter and Rinderle-Ma (2018) Parsing Identifying tokens in the input and analysing the relationship between themAlshemali and Kalita (2020) andChowdhary (2020) Dialogue generation Interacting with human or machine users by understanding and analysing the input and generating responsesChowdhary (2020) Pattern matching Correlating and matching similarities among distinct input messagesChowdhary (2020) andKang et al (2020) Sentiment analysis Analysing the aggregate inputs and inferring the tone and sentiment of the source for generating the input messagesKang et al (2020) Speech recognition Understanding the meaning of what is being conveyed or identifying the source of input from its speechKang et al (2020) andSadat et al (2019) …”