Effects of Differential Privacy and Data Skewness on Membership Inference Vulnerability

@article{Truex2019EffectsOD,
  title={Effects of Differential Privacy and Data Skewness on Membership Inference Vulnerability},
  author={S. Truex and Ling Liu and M. Gursoy and Wenqi Wei and L. Yu},
  journal={2019 First IEEE International Conference on Trust, Privacy and Security in Intelligent Systems and Applications (TPS-ISA)},
  year={2019},
  pages={82-91}
}
  • S. Truex, Ling Liu, +2 authors L. Yu
  • Published 2019
  • Computer Science, Mathematics
  • 2019 First IEEE International Conference on Trust, Privacy and Security in Intelligent Systems and Applications (TPS-ISA)
  • Membership inference attacks seek to infer the membership of individual training instances of a privately trained model. This paper presents a membership privacy analysis and evaluation system, MPLens, with three unique contributions. First, through MPLens, we demonstrate how membership inference attack methods can be leveraged in adversarial ML. Second, we highlight with MPLens how the vulnerability of pre-trained models under membership inference attack is not uniform across all classes… CONTINUE READING

    Figures, Tables, and Topics from this paper.

    LDP-Fed: federated learning with local differential privacy
    • 5
    • PDF
    Adversarial Training with Stochastic Weight Average
    Machine Learning Systems for Intelligent Services in the IoT: A Survey

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 46 REFERENCES
    Membership Inference Attacks Against Machine Learning Models
    • 745
    • PDF
    Privacy Risk in Machine Learning: Analyzing the Connection to Overfitting
    • 140
    Demystifying Membership Inference Attacks in Machine Learning as a Service
    • 27
    Evaluating Differentially Private Machine Learning in Practice
    • 65
    • Highly Influential
    • PDF