4.Privacy Protection Flashcards
main vision of modern healthcare regarding data use
predictive, preventive, personalized, and participatory approaches, emphasizing individual biological conditions and integration of data from various sources like wearables and apps.
challenges in privacy protection for medical data?
conflict between sharing/publishing data and the need to protect individual privacy, alongside complex regulatory requirements.
What is the difference between personal and anonymous data under GDPR?
Personal data relates to identified/identifiable individuals and needs protection, while anonymous data doesn’t relate to identifiable individuals and doesn’t need protection.
How can de-identified genetic data be attacked for re-identification?
By matching it against genetic data in genealogic databases, potentially identifying individuals sharing surnames and Y-chromosome patterns.
What are the common methods for data anonymization?
removal, generalization, aggregation, and perturbation like noise addition.
What is the HIPAA Safe Harbor method?
It’s a heuristic method for anonymizing data by removing or altering 18 types of attributes, ensuring de-identified information cannot be linked to individuals.
it can be vulnerable to re-identification attacks using diagnosis codes due to the dimensionality and sparseness of data.
What is k-anonymity in data privacy?
It’s a model where each record in a dataset has at least k-1 ‘twins’, making it indistinct over quasi-identifier attributes.
: What are the types of scenarios in data anonymization?
Non-interactive, where primary data is modified, and interactive, where query results are modified.
What is data pseudonymization?
it involves processing personal data so that it can no longer be attributed to a specific individual without additional information.
rganizational and Legal Aspects of Privacy Protection:
GDPR exempts anonymous data from data protection principles, while BDSG requires pseudonymization of research data.
Anonymization:
Entails various methods like removal, generalization, aggregation, and perturbation to reduce data uniqueness.
Heuristic methods use simple rules for data modification, while computational approaches quantify privacy risks for optimization.
Privacy models like k-anonymity ensure that each record has at least k-1 ‘twins’ in the dataset, preventing individual identification.
Pseudonymization:
Involves processing personal data so it cannot be attributed to a specific individual without additional information.
It distinguishes identifying information from non-identifying, and is simpler but weaker than anonymization.
Challenge: Data Privacy vs. Data Quality:
Data modification for privacy protection can significantly impact data quality, affecting its statistical properties and analytical validity.
Balancing privacy risk reduction and maintaining data utility is a key challenge in medical research.
