Returns \(\varepsilon = \alpha \Delta^2 / (2\sigma^2)\), the
Renyi differential privacy budget at order \(\alpha\) of adding
\(N(0, \sigma^2)\) noise to a query of \(\ell_2\) sensitivity
\(\Delta\) (Mironov 2017, Proposition 7 and Corollary 3). The
budget curve is linear in \(\alpha\), which is what makes RDP
compose by simple addition.
Usage
morie_rdpc(alpha, sigma, sensitivity = 1)
Arguments
- alpha
Renyi order, strictly greater than 1.
- sigma
Noise standard deviation, positive.
- sensitivity
L2 sensitivity of the query, positive.
Value
A list with epsilon_rdp, estimate,
alpha, sigma, sensitivity, method.
References
Mironov, I. (2017). Renyi differential privacy. IEEE
Computer Security Foundations Symposium, 263-275.
Examples
morie_rdpc(alpha = 2, sigma = 1)$epsilon_rdp # 2 * 1 / 2
#> [1] 1
