# US 2021/0279364 A1 · Data privacy using a podium mechanism

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0279364 A1
- Document ID: US20210279364A1
- Status: Published application
- Kind: A1
- Application number: 17/327,899
- Filed: 2021-05-24
- Published: 2021-09-09
- Earliest claimed: 2019-02-28
- Assignees: Snap Inc.
- Applicants: Snap Inc.
- Inventors: Vasyl Pihur
- Web page: https://patentgrep.com/patent/US20210279364A1

## CPC

- G06F 21/6245: {Protecting personal data, e.g. for financial or medical purposes}
- G06F 17/18: for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)}
- H04L 9/0894: {Escrow, recovery or storing of secret information, e.g. secret key escrow or cryptographic key storage}
- H04L 2209/08: Randomization, e.g. dummy operations or using noise
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Systems and methods are provided for performing operations comprising: storing a set of input data; generating a noise distribution based on a two-step function, wherein a height of the two-step function is determined by a privacy parameter, a width of the two-step function is determined by minimizing a variance of the noise distribution, and wherein a mean of the two-step function is determined by a value of the set of input data to be privatized; applying the noise distribution to the set of input data to generate privatized noisy output data; and transmitting the resulting privatized noisy output data in response to a request for a portion of, or a complete set of, the input data.

## Claims

### Claim 1

A method comprising:
storing, by one or more processors, a set of input data;
generating, by the one or more processors, a noise distribution based on a two-step function;
receiving, from a messaging application, a request to access the set of input data;
selecting a mode of the noise distribution between an exact mode and an approximation mode for evaluating an unconstrained parameter of the two-step function, the exact mode providing a value of the unconstrained parameter from a list of predetermined values of the unconstrained parameter, the approximation mode providing the value for the unconstrained parameter as a function of an exponential function and a constant;
applying the noise distribution according to the selected mode to the set of input data to generate privatized noisy output data to prevent exposure to particular data points representing user information in the set of input data, the privatized noisy output data comprises random data added to the set of input data; and
transmitting, by the one or more processors, the privatized noisy output data.

### Claim 2 (depends on 1)

The method of claim 1, wherein:
a height of the two-step function is determined by a privacy parameter, a width w of the two-step function is determined by minimizing a variance of the noise distribution, and a mean of the two-step function is determined by a value of the set of input data to be privatized.

### Claim 3 (depends on 2)

The method of claim 2, wherein generating the noise distribution comprises:
computing a multiplicative margin m of a spread Δ of the set of input data;
computing the width w of the two-step function as a function of the spread of the set of input data, the multiplicative margin m and an unconstrained parameter s; and
computing a distribution parameter d as a function of the spread of the set of input data, the multiplicative margin m, the unconstrained parameter s, and the privacy parameter.

### Claim 4 (depends on 1)

The method of claim 1, further comprising:
receiving, from a messaging application, a request to access the set of input data; and
transmitting the privatized noisy output data to the messaging application in response to the request.

### Claim 5 (depends on 1)

The method of claim 1, further comprising retrieving, when the selected mode corresponds to the exact mode, a previously stored unconstrained parameter s of the noise distribution, the previously stored unconstrained parameter being previously computed based on a privacy parameter.

### Claim 6 (depends on 1)

The method of claim 1, further comprising computing, when the selected mode corresponds to the approximation mode, an unconstrained parameter s of the noise distribution as a fraction of a privacy parameter.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
receiving a request for a given data x of the set of input data;
computing a location of the two-step function based on the given data x, a width w, a distribution parameter d, and a privacy parameter;
computing probability of a first component p1 as a function of the distribution parameter d, the location, a spread of the set of input data, and a multiplicative margin m; and
computing probability of a second component p2 as a function of the distribution parameter d, the width w, and the privacy parameter.

### Claim 8 (depends on 1)

The method of claim 1, wherein a multiplicative margin m of the noise distribution is computed in accordance with:
m
=
1
+
e
s
+
e
ϵ
+
e
ϵ
-
s
e
ϵ
-
1
where ε is a privacy parameter, and wherein e is an exponential function.

### Claim 9 (depends on 8)

The method of claim 8, wherein a width w of the noise distribution is computed in accordance with:
W
=
Δ
⁢
⁢
m
1
+
e
s
where s is the unconstrained parameter, and A is a spread of the set of input data.

### Claim 10 (depends on 9)

The method of claim 9, wherein a distribution parameter d of the noise distribution is computed in accordance with:
d
=
(
1
+
e
-
s
)
⁢
(
1
+
e
s
)
Δ
⁢
⁢
m
⁡
(
1
+
e
s
+
e
ϵ
+
e
ϵ
-
s
)
.

### Claim 11 (depends on 10)

The method of claim 10, wherein a location t of the noise distribution is computed in accordance with:
t
=
2
⁢
x
-
w
2
⁢
d
⁡
(
e
ϵ
-
1
)
2
⁢
w
⁢
d
⁡
(
e
ϵ
-
1
)
where x is the input value.

### Claim 12 (depends on 11)

The method of claim 11, wherein a probability of a first component p1 is computed in accordance with:
p
1
=
d
⁡
(
t
+
Δ
⁢
⁢
m
2
)
.

### Claim 13 (depends on 11)

The method of claim 11, wherein a probability of a second component p2 is computed in accordance with:
p2=deϵw.

### Claim 14 (depends on 1)

The method of claim 1, further comprising:
in response to generating the privatized noisy output data, selecting from which component of the two-step function to draw noise by generating a uniform random variable and determining which component the random variable lies in.

### Claim 15 (depends on 14)

The method of claim 14, further comprising:
if a first component is randomly selected with a first probability, computing the privatized noisy output data as a first uniform random variable in a range defined as
[
-
Δ
⁢
⁢
m
2
,
t
]
,
where t is a location of the two-step function, Δ is a spread of the set of input data, and m is a multiplicative margin of the noise distribution;
if a second component is randomly selected with a second probability, computing the privatized noisy output data as a second uniform random variable in a range defined as [t, t+w), where w is a width of the noise distribution; and
otherwise, computing the privatized noisy output data as a third uniform random variable in a range defined as
[
t
+
w
,
Δ
⁢
2
2
]
.

### Claim 16

A system comprising:
a processor; and
a memory storing instructions that, when executed by the processor, configure the processor to perform operations comprising:
storing a set of input data;
generating a noise distribution based on a two-step function;
receiving, from a messaging application, a request to access the set of input data;
selecting a mode of the noise distribution between an exact mode and an approximation mode for evaluating an unconstrained parameter of the two-step function, the exact mode providing a value of the unconstrained parameter from a list of predetermined values of the unconstrained parameter, the approximation mode providing the value for the unconstrained parameter as a function of an exponential function and a constant;
applying the noise distribution according to the selected mode to the set of input data to generate privatized noisy output data to prevent exposure to particular data points representing user information in the set of input data, the privatized noisy output data comprises random data added to the set of input data; and
transmitting the privatized noisy output data.

### Claim 17 (depends on 16)

The system of claim 16, wherein:
a height of the two-step function is determined by a privacy parameter, a width w of the two-step function is determined by minimizing a variance of the noise distribution, and a mean of the two-step function is determined by a value of the set of input data to be privatized.

### Claim 18 (depends on 17)

The system of claim 17, wherein generating the noise distribution comprises:
computing a multiplicative margin m of a spread Δ of the set of input data;
computing the width w of the two-step function as a function of the spread of the set of input data, the multiplicative margin m and an unconstrained parameter s; and
computing a distribution parameter d as a function of the spread of the set of input data, the multiplicative margin m, the unconstrained parameter s, and the privacy parameter.

### Claim 19 (depends on 16)

The system of claim 16, wherein the operations further comprise:
retrieving, when the selected mode corresponds to the exact mode, a previously stored unconstrained parameter s of the noise distribution, the previously stored unconstrained parameter being previously computed based on a privacy parameter; and
computing, when the selected mode corresponds to the approximation mode, an unconstrained parameter s of the noise distribution as a fraction of a privacy parameter.

### Claim 20

A non-transitory machine-readable storage medium that includes instructions that, when executed by one or more processors of a machine, cause the machine to perform operations comprising:
storing a set of input data;
generating a noise distribution based on a two-step function;
receiving, from a messaging application, a request to access the set of input data;
selecting a mode of the noise distribution between an exact mode and an approximation mode for evaluating an unconstrained parameter of the two-step function, the exact mode providing a value of the unconstrained parameter from a list of predetermined values of the unconstrained parameter, the approximation mode providing the value for the unconstrained parameter as a function of an exponential function and a constant;
applying the noise distribution according to the selected mode to the set of input data to generate privatized noisy output data to prevent exposure to particular data points representing user information in the set of input data, the privatized noisy output data comprises random data added to the set of input data; and
transmitting the privatized noisy output data.
