Relation Extraction

Library to create vertices and edges from a document annotated with cyber-entity labels and a list of relationship patterns between these cyber entities.

Dependency

The Relation Extraction library shares object models with the Entity Extraction library. To install the dependency locally:

mvn scm:checkout -Dmodule.name=entity-extractor
cd entity-extractor
mvn clean install

Entity Types

Software
- Vendor
- Product
- Version
File
- Name
Function
- Name
Vulnerability
- Name
- Description
- CVE
- MS

Relationship Types

ExploitTargetRelatedObservable Edge

  Exploit Target (e.g. vulnerability) --> Observable (e.g. software)

Sub-Observable Edge

  Observable (e.g. software) --> Observable (e.g. file)

Software, File, Function, Vulnerability Vertices

  Software/file/function/vulnerability properties are part of the same vertex
  
  Example: "... **MS15-035**, which addresses a **remote code execution** bug ..."
  "MS15-035" is extracted as a vulnerability MS property, and "remote code execution" is extracted as a vulnerability description property. This type of relationship indicates that both properties are describing the same vulnerability object.

Relationship Patterns

These patterns are defined in a JSON-formatted file. There are three types of patterns we will need to find:

The exact-match pattern consists of at least two cyber-entity labels, and words or part-of-speech (pos) tags. The annotated document must match the pattern exactly with no extra words.
```
  Example Text: “Tomcat from Apache”
  Pattern: <SW.PRODUCT>, "from", <SW.VENDOR>
```
The fuzzy-match pattern begins and ends with cyber-entity labels, but it allows for extra words to be present in the document that are not reflected in the pattern. The words, or pos tags, in the pattern must appear in the annotated document, but there can be extra words as well. The only restriction is the two entities are no more than 10 words apart in the document.
```
  Example Text: “Apache Tomcat’s latest update 8.0.22”
  Pattern: <SW.PRODUCT>, “update”, <SW.VERSION>
```
The parse-tree-path pattern defines a path through a sentence’s parse tree, where the first and last elements of the path are labeled cyber entities.
```
  Example Parse Tree: (NP (NP (NNP Apache) (NNP Tomcat) (POS 's)) (JJS latest))
  Pattern: <SW.VENDOR> -- NNP -- NP -- NNP -- <SW.PRODUCT>
```

The majority of the knowledge graph's ontology is defined within this file, so the file can be modified while the extractor's mechanics remain the same. The file format is as follows:

{"Patterns": [
{"edgeType": "ExploitTargetRelatedObservable", "patternType": "ExactPattern", "patternSequence": [{"class": "CyberEntity", "value": "sw.product", "vType": "inV"}, {"class": "Token", "value": "update"}, {"class": "Token", "value": "-LRB-"}, {"class": "CyberEntity", "value": "vuln.name", "vType": "outV"}]},
{"vertexType": "software", "patternType": "ExactPattern", "patternSequence": [{"class": "Token", "value": "versions"}, {"class": "Token", "value": "of"}, {"class": "CyberEntity", "value": "sw.product"}, {"class": "Token", "value": "are"}, {"class": "CyberEntity", "value": "sw.version"}, {"class": "POS", "value": "IN"}, {"class": "CyberEntity", "value": "sw.version"}]},
{"vertexType": "software", "patternType": "ParseTreePattern", "patternSequence": [{"class": "CyberEntity", "value": "sw.product"}, {"class": "TreeElement", "value": "NNP"}, ...]},
...
] }

Input

Output from the Entity-Extractor as an Annotation object, which represents the sentences, list of words from the text, along with each word's part of speech tag and cyber domain label.
The String name of the document's source
A JSON file containing the set of patterns to find in the annotated document

Current Process

For each pattern:
- Search the document for candidate instances and for each instance:
  - If the pattern type is ExactPattern, then all tokens listed in the pattern sequence must match
  - If the pattern type is FuzzyPattern, then all the cyber entities must match those in the pattern sequence with no more than 10 words between them, but there can be other tokens in the document that are not represented in the pattern sequence
  - If the pattern type is ParseTreePattern, then the pattern sequence represents a path, between cyber entities, in the parse tree and all elements must match exactly

Output

A JSON-formatted subgraph of the vertices and edges is created, which loosely resembles the STIX data model

 { "vertices" : 
 	{
 		id1 : { "vertexType": "software", "vendor": "Apache", "product": "Tomcat"},
 		id2 : { "vertexType": "vulnerability", "description": "buffer overflow"},
 		...
 	},
   "edges" : 
   	[
   		{"id": "vuln_Tomcat_1234", "inV": "id1", "outV": "id2", "label": "ExploitTargetRelatedObservable"},
 		...
   	]
 }

Usage

EntityLabeler labeler = new EntityLabeler();
Annotation doc = labeler.getAnnotatedDoc("My Doc", exampleText);
RelationExtractor rx = new RelationExtractor();
String graph = rx.createSubgraph(doc, "My source");

Test

Install the dependency as described above
Run the commands:
```
 mvn clean package
 mvn test
```

License

This software is freely distributable under the terms of the MIT License.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS, THE U.S. GOVERNMENT, OR UT-BATTELLE BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

stucco/relation-extractor