privacy in pharmacogenetics
TRANSCRIPT
PRIVACY IN PHARMACOGENETICS AN END-TO-END CASE STUDY OF
PERSONALIZED WARFARIN DOSINGA REPORT WRITTEN BY:
AL ANOUD ALQOUFI- SHATAHA AL TALHI
KING SAUD UNIVERSITY
IS 563 INFORMATION SECURITY MANAGEMENT &
AUDIT
SUPERVISED BY : DR. MOHAMMED ALHUSSEIN
Outline
• Introduction
• Pharmacogenetic and Warfarin
• Model inversion
• Differential privacy
• Results
• Conclusion
• References
Introduction
A case study was introduced using warfarin dosage which is the most used in pharmacogenetic modeling. researchers emphasized on the
amount of private information that might be revealed by model inversion attacks .
Differential privacy was used for building pharmacogenetic models
The case study concluded that differential privacy cannot grantee privacy protection without effecting clinical efficiency.
Pharmacogenetic & Warfarin
Warfarin It is the most popular anticoagulant used to decrease the possibility of heart attacks and strokes to occur .
Low dose
Pharmacogenetic = Pharma + Genetic
Pharmacogenetic & Warfarin
Genotype
Clinical variables
Trained model
Medical Guidance
Population dataset
Trained model
Learning Algorithm
Linear Model f(x)
Sqrt(dose) = 5.6044 + 0.2614 * age + 0.8677 * vkorc1=A/G - 1.6974 * vkorc1=A/A - 1.9206 * cyp2c9=*2/*3 - 2.3312 * cyp2c9=*3/*3 + 0.1092 * asian race - 0.2760 * black or african american - ……
Model Inversion
Differential privacy
DP objective is to prevent attackers to conclude that a subject was in the set used to construct a model or not.
Most DP mechanisms “Add noise” according to privacy budget .
DP Insures patients’ privacy
Results
CONCLUSION
Current methods fail to balance privacy and utility in Pharmacogenetic models
REFERENCES 1. M. Fredrikson, E. Lantz and S. Jha, "Privacy in pharmacogenetics: An end-to-end case study of personalized
warfarin dosing," in 23rd USENIX Security Symposium (USENIX Security 14), San Diego, 2014.
2. J. A. Johnson and d. L. H. Cavallari, "Warfarin pharmacogenetics," Trends in cardiovascular medicine 25, vol. 1, pp. 33-41, 2015.
3. A. Tarantola, Inverse Problem Theory and Methods for Model Parameter Estimation, Paris: Siam, 2005.
4. D. F. Kamal and K. E. Emam, "The application of differential privacy to health data.," in Proceedings of the 2012 Joint EDBT/ICDT Workshops.ACM, New York, 2012. 5. J. Zhang, Z. Zhang, X. Xiao, Y. Yang, and M. Winslett. Functional mechanism: regression analysis under differential privacy. In VLDB, 2012.6. S. Vinterbo. Differentially private projected histograms: Construction and use for prediction. In ECML-PKDD, 2012.7. D. Hand and R. Till. A simple generalization of the area under the ROC curve for multiple class classification problems. Machine Learning, 45(2):171– 186, 2001.