Privacy Enhancing Techniques

Dated Jan 22, 2019; last modified on Sun, 14 Mar 2021

Examples include: secure communication, anti-tracking features in browsers and mobile OSes, secure multiparty computation, differential privacy, anonymous communication, etc.

		Random Link ¯\_(ツ)_/¯
Jan 22, 2019	»	Differential Privacy (2 items) Applications for DP; Building Intuition About DP;
Jan 22, 2019	»	Secure Multiparty Computation 3 min; updated Oct 27, 2021 Timeline of Secure Multi-Party Communication In 1982, secure two-party computation (2PC) was introduced for problems that are boolean predicates, e.g. Yao’s Millionaires' Problem that asks whether \(a \ge b\) is true without revealing the actual values of \(a\) and \(b\). Andrew Yao generalized 2PC for any feasible computation in 1986. Goldreich, Micali and Wigderson later generalized it to secure multiparty communication. Yao-based protocols requires that the function to be securely evaluated be represented as a circuit, but an efficient transformation is not trivial. The Fairplay system transforms a program in a high-level language into a boolean circuit representation, garbles that circuit and then executes a protocol to evaluate the garbled circuit. AES’s circuit has \(\approx 50K\) gates, while the 4095-bit edit distance function has \(\approx 6B\) gates. ...
Jan 23, 2021	»	Research on Privacy Enhancing Techniques 2 min; updated Feb 12, 2023 Journals note that prediction services can still make accurate predictions using a fraction of the data collected from a user device. They propose Cloak, which suppresses non-pertinent features (i.e. those features which can consistently tolerate addition of noise without degrading utility) to the prediction task. Cloak has a provable degree of privacy, and unlike cryptographic techniques, does not degrade prediction latency. Using the training data, labels, a pre-trained model and a privacy-utility knob, they (1) find the pertinent features through perturbation training, and (2) learn utility-preserving constant values for suppressing the non-pertinent data. During the inference phase, they efficiently compute a sifted representation which is sent to the service provider. ...