More Like this

1. MIST: Defending Against Membership Inference Attacks Through Membership-Invariant Subspace Training

   Li, J; Li, N; Ribeiro, B (August 2024, USENIX)
2. Survival Mixed Membership Blockmodel

   https://doi.org/10.1080/01621459.2023.2213466  

   Song, Fangda; Chu, Jing; Ma, Shuangge; Wei, Yingying (April 2024, Journal of the American Statistical Association)
3. Credibility in Private Set Membership

   https://doi.org/10.1007/978-3-031-31371-4_6  

   Garg, S.; Hajiabadi, M.; Jain, A.; Jin, Z.; Pandey, O.; Shiehian, S. (May 2023, Public-Key Cryptography (PKC 2023))

   Boldyreva, A.; Kolesnikov, V. (Ed.)

   A private set membership (PSM) protocol allows a “receiver” to learn whether its input x is contained in a large database 𝖣𝖡 held by a “sender”. In this work, we define and construct credible private set membership (C-PSM) protocols: in addition to the conventional notions of privacy, C-PSM provides a soundness guarantee that it is hard for a sender (that does not know x) to convince the receiver that 𝑥∈𝖣𝖡. Furthermore, the communication complexity must be logarithmic in the size of 𝖣𝖡. We provide 2-round (i.e., round-optimal) C-PSM constructions based on standard assumptions: We present a black-box construction in the plain model based on DDH or LWE. Next, we consider protocols that support predicates f beyond string equality, i.e., the receiver can learn if there exists 𝑤∈𝖣𝖡 such that 𝑓(𝑥,𝑤)=1. We present two results with transparent setups: (1) A black-box protocol, based on DDH or LWE, for the class of NC1 functions f which are efficiently searchable. (2) An LWE-based construction for all bounded-depth circuits. The only non-black-box use of cryptography in this construction is through the bootstrapping procedure in fully homomorphic encryption. As an application, our protocols can be used to build enhanced round-optimal leaked password notification services, where unlike existing solutions, a dubious sender cannot fool a receiver into changing its password. https://doi.org/10.1007/978-3-031-31371-4_6 

   more » « less
4. Minimum Membership Covering and Hitting

   https://doi.org/10.1016/j.tcs.2021.05.002  

   Mitchell, Joseph S.B.; Pandit, Supantha (July 2021, Theoretical Computer Science)

   null (Ed.)
5. Revisiting Membership Inference Under Realistic Assumptions

   https://doi.org/10.2478/popets-2021-0031  

   Jayaraman, Bargav; Wang, Lingxiao; Knipmeyer, Katherine; Gu, Quanquan; Evans, David (January 2021, Proceedings on Privacy Enhancing Technologies)

   Abstract We study membership inference in settings where assumptions commonly used in previous research are relaxed. First, we consider cases where only a small fraction of the candidate pool targeted by the adversary are members and develop a PPV-based metric suitable for this setting. This skewed prior setting is more realistic than the balanced prior setting typically considered. Second, we consider adversaries that select inference thresholds according to their attack goals, such as identifying as many members as possible with a given false positive tolerance. We develop a threshold selection designed for achieving particular attack goals. Since previous inference attacks fail in imbalanced prior settings, we develop new inference attacks based on the intuition that inputs corresponding to training set members will be near a local minimum in the loss function. An attack that combines this with thresholds on the per-instance loss can achieve high PPV even in settings where other attacks are ineffective. 

   more » « less