Detection of spam using contextual analysis of data sources

• US 8,909,591 B1
• Filed: 04/16/2014
• Issued: 12/09/2014
• Est. Priority Date: 07/26/2011
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Grundman, Douglas Richard, Yuksel, Baris, Adarsh, Anurag, Janawadkar, Piyush
• Primary Examiner(s): Chen, Alan
• Assistant Examiner(s): Chubb, Mikayla
• Application Number:
  US14/254,335
• Time in Patent Office: 237 Days
• Field of Search: None
• US Class Current: 706/52
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06Q 10/107 Computer-aided management of electronic mailing [e-mailing]
• CPC Class Codes: H04L 63/20 for managing network security; network security policies in general filtering policies H04L63/0227

Abstract

Aspects of the disclosure provide for detection of spam business listings. Aspects operate to identify business listing characteristics in trusted sources and untrusted sources. As untrusted sources are likely to contain more spam, characteristics that are present in untrusted sources but not present in trusted sources are typically indicative of spam listings, and vice versa. Thus, statistical analysis of the frequency of characteristics within each source may be used to identify common characteristics of spam listings. These characteristics may further be analyzed in specific listing contexts, as different listing contexts (e.g., different types of businesses) typically use different terms and vocabularies, such that terms that are indicative of spam in one context may not be indicative of spam in another. Various methods for leveraging this context-specific statistical information to improve spam detection operations are disclosed.

Claims

1. A computer implemented method for identifying business listings, the method comprising:

determining, using one or more processors, a first frequency value of a business listing characteristic within a first plurality of business listings received from a first source, the first plurality of business listings being associated with a particular business listing context;

determining, using the one or more processors, a second frequency value of the business listing characteristic within a second plurality of business listings received from a second source, the second plurality of business listings being associated with the particular business listing context;

determining, using the one or more processors, a frequency differential between the first frequency value and the second frequency value;

in response to the frequency differential exceeding a threshold differential, identifying, using the one or more processors, the business listing characteristic as a differential characteristic; and

identifying, using the one or more processors, a particular business listing of the plurality of business listings as a spam listing using the differential characteristic.

2. The method of claim 1, wherein the business listing context is a type of business.

3. The method of claim 1, wherein the threshold differential is a standard deviation value.

4. The method of claim 1, wherein the business listing characteristic is at least one of a title length, a text term, a phone number, and an address.

5. The method of claim 1, wherein identifying the particular business listing further comprises:

identifying, using the one or more processors, a spam score for the particular business listing using the differential characteristic; and

wherein identifying the past business listing as the spam listing is done in response to the spam score being greater than a threshold.

6. The method of claim 5, wherein identifying the particular business listing further comprises:

mapping, using the one or more processors, the spam score to a context-specific curve associated with the business listing context to determine a probability that the particular business listing is spam; and

wherein identifying the past business listing as the spam listing is done in response to the probability being greater than a probability threshold.

7. The method of claim 6, wherein the probability threshold is a context-specific probability threshold associated with the business listing context.

8. The method of claim 1, wherein the first source is a trusted source and the second source is an untrusted data source.

9. A processing system for identifying business listings, the processing system comprising:

one or more computing devices, each of the one or more computing devices having one or more processors; and

a memory, coupled to the one or more processors, for storing business listings;

wherein the one or more computing devices is configured to;

determine a first frequency value of a business listing characteristic within a first plurality of business listings received from a first source, the first plurality of business listings being associated with a particular business listing context;

determine a second frequency value of the business listing characteristic within a second plurality of business listings received from a second source, the second plurality of business listings being associated with the particular business listing context;

determine a frequency differential between the first frequency value and the second frequency value;

in response to the frequency differential exceeding a threshold differential, identify the business listing characteristic as a differential characteristic; and

identify a particular business listing of the plurality of business listings as a spam listing using the differential characteristic.

10. The processing system of claim 9, wherein the business listing context is a type of business.

11. The processing system of claim 9, wherein the threshold differential is a standard deviation value.

12. The processing system of claim 9, wherein the business listing characteristic is at least one of a title length, a text term, a phone number, and an address.

13. The processing system of claim 9, wherein the one or more processors is further configured to:

identify a spam score for the particular business listing using the differential characteristic; and

identify the particular business listing as the spam listing in response to the spam score being greater than a threshold.

14. The processing system of claim 13, wherein the one or more processors is further configured to:

map the spam score to a context-specific curve associated with the business listing context to determine a probability that the particular business listing is spam; and

identify the business listing as a spam listing in response to the probability being greater than a probability threshold.

15. The processing system of claim 14, wherein the probability threshold is a context-specific probability threshold associated with the business listing context.

16. The processing system of claim 9, wherein the first source is a trusted source and the second source is an untrusted data source.

17. A non-transitory, computer-readable medium on which instructions are stored, the instructions, when executed by one or more computing devices performs a method, the method comprising:

determining a first frequency value of a business listing characteristic within a first plurality of business listings received from a first source, the first plurality of business listings being associated with a particular business listing context;

determining a second frequency value of the business listing characteristic within a second plurality of business listings received from a second source, the second plurality of business listings being associated with the particular business listing context;

determining a frequency differential between the first frequency value and the second frequency value;

in response to the frequency differential exceeding a threshold differential, identifying the business listing characteristic as a differential characteristic; and

identifying a particular business listing of the plurality of business listings as a spam listing using the differential characteristic.

18. The non-transitory, computer-readable medium of claim 17, wherein the business listing context is a type of business.

19. The non-transitory, computer-readable medium of claim 17, wherein the threshold differential is a standard deviation value.

20. The non-transitory, computer-readable medium of claim 17, wherein the business listing characteristic is at least one of a title length, a text term, a phone number, and an address.