# US 2012/0254388 A1 · Method and apparatus for communications analysis

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

- Number: US 2012/0254388 A1
- Document ID: US20120254388A1
- Status: Published application
- Kind: A1
- Application number: 13/365,760
- Filed: 2012-02-03
- Published: 2012-10-04
- Earliest claimed: 2011-02-03
- Assignees: Roke Manor Research Limited
- Inventors: Neil Duxbury; David Findlay
- Web page: https://patentgrep.com/patent/US20120254388A1

## CPC

- G06F 16/285: {Clustering or classification}
- H04L 43/026: using flow identification
- Y02D 30/50: in wire-line communication networks, e.g. low power modes or reduced link rate
- 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)
- Y02D: CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE

## Abstract

A method of grouping communication sessions, the method comprising: selecting a plurality of communications sessions from a data stream; determining which data structures, of said communication sessions, occur more frequently than chance; and sorting the communication sessions into groups, wherein communication sessions which have similar data structures, determined to occur more frequently than chance, are sorted into the same group.

## Claims

### Claim 1

A method of grouping communication sessions, the method comprising:
selecting a plurality of communications sessions from a data stream;
determining which data structures, of said communication sessions, occur more frequently than chance; and
sorting the communication sessions into groups, wherein communication sessions which have similar data structures, determined to occur more frequently than chance, are sorted into the same group.

### Claim 2 (depends on 1)

A method according to claim 1, further comprising: searching each communication session for at least one entity, after said sessions have been selected; and extracting each occurrence of an entity and a context of the entity; wherein said step of determining is done based on data structures within said contexts.

### Claim 3 (depends on 2)

A method according to claim 2, wherein a context is a predetermined number of characters before and/or after each entity.

### Claim 4 (depends on 1)

A method according to claim 1, further comprising generating a plurality of data structures from each context.

### Claim 5 (depends on 4)

The method of claim 4, further comprising generating a vector to represent each communication session, wherein the data structures are the vector components.

### Claim 6 (depends on 5)

The method of claim 5, further comprising using a similarity measure to determine the distance between session vectors.

### Claim 7 (depends on 6)

The method of claim 6, wherein said sessions are sorted into groups using a clustering algorithm.

### Claim 8 (depends on 2)

The method of claim 2, wherein a data structure of a context is a sub-sequence of characters within a context.

### Claim 9 (depends on 8)

The method of claim 8, wherein said data structures are a plurality of overlapping ngrams.

### Claim 10 (depends on 9)

The method of claim 9, wherein an ngram is n consecutive characters, and each ngram overlaps by n−1.

### Claim 11 (depends on 9)

The method of claim 9, further comprising plotting a histogram of ngram session frequencies.

### Claim 12 (depends on 11)

The method of claim 11, further comprising identifying ngrams which occur more frequently than chance.

### Claim 13 (depends on 1)

The method of claim 1, wherein the step of selecting sessions includes sampling packets of data from the data stream and reconstructing sessions from the packets.

### Claim 14

An apparatus comprising:
a computer; and
a computer readable medium having computer code stored thereon, that when executed by the computer, causes the apparatus to:
select a plurality of communications sessions from a data stream;
determine which data structures, of said communication sessions, occur more frequently than chance; and
sort the communication sessions into groups, wherein communication sessions which have similar data structures, determined to occur more frequently than chance, are sorted into the same group.

### Claim 15

(canceled)

### Claim 16

A computer readable medium having code stored thereon which, when run on a computer, causes the computer to:
select a plurality of communications sessions from a data stream;
determine which data structures, of said communication sessions, occur more frequently than chance; and
sort the communication sessions into groups, wherein communication sessions which have similar data structures, determined to occur more frequently than chance, are sorted into the same group.

### Claim 17

A method of grouping communications sessions, the method comprising:
extracting a plurality of communication sessions from a data stream, each communication session comprising a sequence of characters;
analysing the communication sessions to determine sequences of characters which exhibit repeatable behaviour; and
sorting communications sessions having similar sequences of characters into groups.

### Claim 18

A method of grouping communication sessions, the method comprising:
extracting a plurality of communications sessions from a data stream;
searching the sessions for at least one entity, the entity being a string of characters;
selecting the entities and the associated entity contexts, a context being a string of characters of a predetermined length before and/or after an entity;
dividing the contexts into a plurality of context elements, a context element being a string of characters which is shorter than the context;
determining which context elements occur more frequently than chance;
representing each communications session using the context elements determined to occur more frequently than chance;
expressing the representation as a vector, where each context element is a component of the vector;
comparing each session using a similarity measure; and
clustering the sessions based on vector similarity.

### Claim 19

A method of configuring a sensor to extract data from a communication data stream, using a group of communication sessions representing a particular communications protocol, the group comprising data structures representative of that protocol, the method comprising:
generating a plurality of records representing said data structures, each record having a particular pattern;
grouping said records based on the similarity of said patterns, such that each group includes records having the same pattern;
generating a template based on the pattern of each group; and
configuring said sensor using said template.

### Claim 20 (depends on 19)

A method according to claim 19, wherein, each record includes an entity and the context of that entity.

### Claim 21 (depends on 19)

A method according to claim 19, further comprising:
extracting said data structures from a group of communication sessions, prior to generating said records;
searching for occurrences of those structures in a data stream;
extracting packets containing those occurrences; and
generating said records on the basis of the extracted packets.

### Claim 22 (depends on 21)

A method according to claim 21, further comprising aligning the records which have been generated.

### Claim 23 (depends on 22)

A method according to claim 22, further comprising determining the similarity of the records after aligning the records.

### Claim 24

(canceled)

### Claim 25 (depends on 22)

A method according to claim 22, wherein the alignment is performed using the Needleman-Wunsch algorithm.

### Claim 26 (depends on 23)

A method according to claim 23, wherein the step of determining similarity is performed using a similarity measure.

### Claim 27 (depends on 26)

A method according to claim 26, wherein the similarity measure is a cosine-like similarity measure.

### Claim 28

An apparatus comprising:
a computer; and
a computer readable medium having computer code stored thereon, that when executed by the computer, causes the apparatus to:
extract a plurality of communications sessions from a data stream;
search the sessions for at least one entity, the entity being a string of characters:
select the entities and the associated entity contexts, a context being a string of characters of a predetermined length before and/or after an entity;
divide the contexts into a plurality of context elements, a context element being a string of characters which is shorter than the context;
determine which context elements occur more frequently than chance:
represent each communications session using the context elements determined to occur more frequently than chance;
express the representation as a vector, where each context element is a component of the vector;
compare each session using a similarity measure; and
cluster the sessions based on vector similarity.

### Claim 29

(canceled)

### Claim 30

A computer readable medium having code stored thereon which, when run on a computer, causes the computer to:
extract a plurality of communications sessions from a data stream;
search the sessions for at least one entity, the entity being a string of characters:
select the entities and the associated entity contexts, a context being a string of characters of a predetermined length before and/or after an entity;
divide the contexts into a plurality of context elements, a context element being a string of characters which is shorter than the context;
determine which context elements occur more frequently than chance;
represent each communications session using the context elements determined to occur more frequently than chance;
express the representation as a vector, where each context element is a component of the vector;
compare each session using a similarity measure; and
cluster the sessions based on vector similarity.

### Claim at position 31

31-34. (canceled)
