Emoji frequency detection and deep link frequency

• US 9,705,908 B1
• Filed: 09/24/2016
• Issued: 07/11/2017
• Est. Priority Date: 06/12/2016
• Status: Expired due to Fees

• Current Assignee: Apple Inc.
• Original Assignee: Apple Inc.
• Inventors: Thakurta, Abhradeep Guha, Vyrros, Andrew H., Vaishampayan, Umesh S., Kapoor, Gaurav, Freudinger, Julien, Legendre, Arnaud, Prakash, Vipul Ved, Duplinsky, Steven
• Primary Examiner(s): Schwartz, Darren B
• Application Number:
  US15/275,360
• Time in Patent Office: 290 Days
• CPC Class Codes: G06F 21/6254 by anonymising data, e.g. decorrelating personal data from the owner's identification
• CPC Class Codes: G06F 40/134 Hyperlinking
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G06F 40/274 Converting codes to words; Guess-ahead of partial word inputs
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: H04L 63/0421 Anonymous communication, i.e. the party's identifiers are hidden from the other party or parties, e.g. using an anonymizer
• CPC Class Codes: H04L 63/1425 Traffic logging, e.g. anomaly detection

Abstract

Systems and methods are disclosed for generating term frequencies of known terms based on crowdsourced differentially private sketches of the known terms. An asset catalog can be updated with new frequency counts for known terms based on the crowdsourced differentially private sketches. Known terms can have a classification. A client device can maintain a privacy budget for each classification of known terms. Classifications can include emojis, deep links, locations, finance terms, and health terms, etc. A privacy budget ensures that a client does not transmit too much information to a term frequency server, thereby compromising the privacy of the client device.

Claims

1. A computer-implemented method practiced on a client device, comprising:

receiving a term from an application on the client device by a differential privacy engine eecuting on the client device;

applying, by the differential privacy engine, a differential privacy algorithm to the term thereby generating a differentially private sketch of the term, wherein generating the differentially private sketch of the term comprises;

determining, by the differential privacy engine, a noise constant c,

initializing, by the differential privacy engine, a vector v having dimension m;

randomly selecting a hash function h from a set of hash functions H and setting v[h(term)]=a constant, c2;

generating a vector b of dimension m with each bit b[j]∈

{−

1, 1} for each j∈

m, and b[j] having a value of 1 with a predetermined probability;

generating the differentially private sketch using noise constant c, constant c2, and vector b;

storing, by the differential privacy engine, the differentially private sketch for transmission to a term frequency server in response to determining, by the differential privacy engine, that there is privacy budget available to transmit the differentially private sketch of the term to the term frequency server.

2. The method of claim 1, wherein generating the noise constant c is dependent upon a privacy constant ε

and differentially private sketch comprises;

initializing the vector v having dimension m with v[i]=0 for each i∈

[m].

3. The method of claim 2, wherein generating the differentially private sketch further comprises:

setting c2=1;

generating a vector v_Hadamard=·

H_m·

v, wherein <

1, and H_m is a Hadamard matri of dimension m;

randomly selecting an inde j∈

[m] of vector b; and

returning, as the differentially private sketch, c·

b·

v_Hadamard[j], the selected hash function h, and the selected inde j.

4. The method of claim 1, further comprising:

associating the term with a classification; and

associating a privacy budget with the classification.

5. The method of claim 1, further comprising:

storing the term in a blacklist storage in response to storing the differentially private sketch for transmission to a term frequency server.

6. The method of claim 1, wherein the noise constant is dependent upon a privacy constant ε

and generating the differentially private sketch further comprises;

initializing the vector v having dimension m with elements v[i]=(−

c)ⁱ for each i∈

[m].

7. The method of claim 6, wherein generating the differentially private sketch further comprises:

setting c2=noise constant c;

returning differentially private sketch

8. A non-transitory computer readable medium programmed with instructions that, when eecuted by a processing system, perform operations, comprising:

receiving a term from an application on the client device by a differential privacy engine eecuting on a client device;

applying, by the differential privacy engine, a differential privacy algorithm to the term thereby generating a differentially private sketch of the term, wherein generating the differentially private sketch of the term comprises;

determining, by the differential privacy engine, a noise constant c, and

initializing, by the differential privacy engine, a vector v having dimension m;

randomly selecting a hash function h from a set of hash functions H and setting v[h(term)]=a constant, c2;

generating a vector b of dimension m with each bit b[j]∈

{−

1, 1} for each j∈

m, and b[j] having a value of 1 with a predetermined probability;

generating the differentially private sketch using noise constant c, constant c2, and vector b;

storing, by the differential privacy engine, the differentially private sketch for transmission to a term frequency server in response to determining that there is privacy budget available to transmit the differentially private sketch of the term to the term frequency server.

9. The medium of claim 8, wherein the noise constant is dependent upon a privacy constant ε

and generating the differentially private sketch further comprises;

initializing the vector v having dimension m with v[i]=0 for each i∈

[m].

10. The medium of claim 9, wherein generating the differentially private sketch further comprises:

setting c2=1;

generating a vector v_Hadamard=·

H_m·

V, wherein <

1, and H_m is a Hadamard matri of dimension m;

randomly selecting an inde j∈

[m] of vector b; and

retuning, as the differentially private sketch, c·

b·

v_Hadamard[j], the selected hash function h, and the selected inde j.

11. The medium of claim 8, the operations further comprising:

associating the term with a classification; and

associating a privacy budget with the classification.

12. The medium of claim 8, the operations further comprising:

storing the term in a blacklist storage in response to storing the differentially private sketch for transmission to a term frequency server.

13. The medium of claim 8, the operations further comprising:

discarding the term in response to determining that the term appears on a blacklist of terms.

14. The medium of claim 8, the operations further comprising:

discarding the term in response to determining one of;

the term appears in a blacklist storage of terms that have had a differentially private sketch for the term previously transmitted to the term frequency server;

or

the term appears in a user private dictionary and the term has been previously indicated by the user that the term a differentially private sketch of the term is not to be transmitted to the term frequency server.

15. A system comprising:

a processing system coupled to a memory programmed with eecutable instructions that, when eecuted by the processing system perform operations, comprising;

receiving a term from an application on a client device, by a differential privacy engine eecuting on the client device;

applying, by the differential privacy engine, a differential privacy algorithm to the term thereby generating a differentially private sketch of the term, wherein generating a differentially private sketch of the term comprises;

determining, by the differential privacy engine, a noise constant c, and

initializing, by the differential privacy engine, a vector v having dimension m;

randomly selecting a hash function h from a set of hash functions H and setting v[h(term)]=a constant, c2;

generating a vector b of dimension m with each bit b[j]∈

{−

1, 1} for each j∈

m, and b[j] having a value of 1 with a predetermined probability;

generating the differentially private sketch using noise constant c, constant c2, and vector b;

storing the differentially private sketch for transmission to a term frequency server in response to determining that there is privacy budget available to transmit the differentially private sketch of the term to the term frequency server.

16. The system of claim 15, wherein the noise constant is dependent upon a privacy constant ε

and generating the differentially private sketch further comprises;

initializing the vector v having dimension m with v[i]=0 for each i∈

[m].

17. The system of claim 16, wherein generating the differentially private sketch further comprises:

setting c2=1;

generating a vector v_Hadamard=·

H_m·

v, wherein <

1, and H_m is a Hadamard matri of dimension m;

randomly selecting an inde j∈

[m] of vector b; and

returning, as the differentially private sketch, c·

b·

v_Hadamard[j], the selected hash function, and the selected inde j.

18. The system of claim 15, the operations further comprising:

associating the term with a classification; and

associating a privacy budget with the classification.

19. The system of claim 15, the operations further comprising:

storing the term in a blacklist storage in response to storing the differentially private sketch for transmission to a term frequency server.

20. The system of claim 15, the operations further comprising:

discarding the term in response to determining that the term appears on a blacklist of terms.

21. The system of claim 15, the operations further comprising:

discarding the term in response to determining one of;

the term appears in a transmitted terms storage of terms that have had a differentially private sketch for the term previously transmitted to the term frequency server;

or

the term appears in a user private dictionary and the term has been previously indicated by the user that the term a differentially private sketch of the term is not to be transmitted to the term frequency server.