Graph analytics in Linkurious Enterprise
0 likes269 views
Linkurious, founded in 2013, offers graph analytics capabilities through Linkurious Enterprise, enabling faster detection and investigation of sophisticated threats. The platform utilizes various analytics approaches, including graph pattern-matching and algorithms, to extract valuable insights from graph data. With a focus on visualization and ad-hoc investigations, Linkurious integrates with Neo4j to enhance analytics for business users and developers.
![Graph
pattern-matching
example for an
alert.
//Identify addresses used to register more than 5 distinct companies
MATCH (a:Address)-[r]-(b)
WITH count(b) as score, a, COLLECT( distinct r) as r, COLLECT(
distinct b) as b
WHERE score>5
RETURN a, r, b, score](https://image.slidesharecdn.com/webinar18linkuriousandgraphanalytics-180822144634/85/Graph-analytics-in-Linkurious-Enterprise-13-320.jpg)
![Graph
pattern-matching
example for a
query template.
//What other officers are connected to the same address or the same
company?
MATCH (a:Officer)
WHERE ID(a) = {{"My param":node}}
WITH a
MATCH
(a)-[rel3:OFFICER_OF|:SHAREHOLDER_OF]->(d:Entity)<-[rel4:OFF
ICER_OF|:SHAREHOLDER_OF]-(e:Officer)
RETURN a, d, e, rel3, rel4](https://image.slidesharecdn.com/webinar18linkuriousandgraphanalytics-180822144634/85/Graph-analytics-in-Linkurious-Enterprise-14-320.jpg)
![PageRank
algorithm used in
an alert.
//Detect French entities with a high PageRank
MATCH (a:Entity)-[r]-(b)
WHERE b.countries = "France"
WITH a.pagerank as score, a, COLLECT( distinct r) as r, COLLECT(
distinct b) as b, count(b) as degree
RETURN a, score, a.name as name, r, b, degree
ORDER BY score DESC](https://image.slidesharecdn.com/webinar18linkuriousandgraphanalytics-180822144634/85/Graph-analytics-in-Linkurious-Enterprise-16-320.jpg)
![Graph algorithm
example:
community
detection.
//Detection of communities with the Louvain algorithm
CALL algo.louvain(
'MATCH (p:Officer) RETURN id(p) as id',
'MATCH
(p1:Officer)-[:OFFICER_OF]->(:Entity)<-[:OFFICER_OF]-(p2:Officer)
RETURN id(p1) as source, id(p2) as target',
{graph:'cypher',write:true});](https://image.slidesharecdn.com/webinar18linkuriousandgraphanalytics-180822144634/85/Graph-analytics-in-Linkurious-Enterprise-17-320.jpg)
![Community
detection used in
a query template.
//Retrieve the nodes who belong to the same community
MATCH (a:Officer)
WHERE ID(a) = {{"My param":node}}
WITH a
MATCH p = (a:Officer)-[*..4]-(b:Officer)
WHERE a.communityLouvain = b.communityLouvain
RETURN p](https://image.slidesharecdn.com/webinar18linkuriousandgraphanalytics-180822144634/85/Graph-analytics-in-Linkurious-Enterprise-18-320.jpg)
Graph analytics in Linkurious Enterprise
- 1. SAS founded in 2013 in Paris | http://linkurio.us | @linkurious Graph analytics in Linkurious Enterprise.
- 2. Why and how to use graph analytics with Linkurious Enterprise.
- 3. Detect and investigate sophisticated threats faster with Linkurious Enterprise.
- 4. Linkurious Enterprise is only for visualization Linkurious Enterprise can be used to make analytics available to end-users Misconception #1
- 5. Analysis of graph data = graph analytics All sort of analytics approaches can be applied to graph data Misconception #2
- 6. Data analytics can accelerate the discovery of interesting information.
- 7. 3 different options Graph pattern-matching Ex: Show all companies connected to at least 2 directors connect to 1 address in the UK Graph algorithms Ex: community detection (Louvain, label propagation), centrality (PageRank, closeness), shortest path Other algorithms Ex: linear regression, machine learning, deep learning 21 3
- 8. Apply an iterative approach to maximize success. Maturity Visualization and ad-hoc investigations Use of graph patterns and algorithms in detection and exploration Use of other data analytics approaches (ML, NLP)
- 9. What it looks like from an architecture perspective. Linkurious Enterprise Graph DB (OLTP) Graph processor (OLAP) 2. Persist results 1. Export data
- 10. Examples based on the Paradise Papers dataset.
- 11. Architecture blueprint for graph analytics with Linkurious Enterprise and Neo4j. Linkurious Enterprise Neo4j + Neo4j Graph Algorithms (plugin) Business users Developers, data scientists
- 12. Graph analytics can be used via filters, sizes, colors, alerts, query templates.
- 13. Graph pattern-matching example for an alert. //Identify addresses used to register more than 5 distinct companies MATCH (a:Address)-[r]-(b) WITH count(b) as score, a, COLLECT( distinct r) as r, COLLECT( distinct b) as b WHERE score>5 RETURN a, r, b, score
- 14. Graph pattern-matching example for a query template. //What other officers are connected to the same address or the same company? MATCH (a:Officer) WHERE ID(a) = {{"My param":node}} WITH a MATCH (a)-[rel3:OFFICER_OF|:SHAREHOLDER_OF]->(d:Entity)<-[rel4:OFF ICER_OF|:SHAREHOLDER_OF]-(e:Officer) RETURN a, d, e, rel3, rel4
- 15. Graph algorithm example: PageRank algorithm. //Computation of PageRank CALL algo.pageRank(null,null,{write:true,writeProperty:'pagerank_g'})
- 16. PageRank algorithm used in an alert. //Detect French entities with a high PageRank MATCH (a:Entity)-[r]-(b) WHERE b.countries = "France" WITH a.pagerank as score, a, COLLECT( distinct r) as r, COLLECT( distinct b) as b, count(b) as degree RETURN a, score, a.name as name, r, b, degree ORDER BY score DESC
- 17. Graph algorithm example: community detection. //Detection of communities with the Louvain algorithm CALL algo.louvain( 'MATCH (p:Officer) RETURN id(p) as id', 'MATCH (p1:Officer)-[:OFFICER_OF]->(:Entity)<-[:OFFICER_OF]-(p2:Officer) RETURN id(p1) as source, id(p2) as target', {graph:'cypher',write:true});
- 18. Community detection used in a query template. //Retrieve the nodes who belong to the same community MATCH (a:Officer) WHERE ID(a) = {{"My param":node}} WITH a MATCH p = (a:Officer)-[*..4]-(b:Officer) WHERE a.communityLouvain = b.communityLouvain RETURN p
- 19. Unlock the value of your graph data with Linkurious Enterprise.
- 20. We work with more than 50+ organizations worldwide.
- 21. The compliance team of BforBank uses graph pattern search to detect fraud.
- 22. Identifying criminal communities based on communication patterns.
- 23. Using machine learning to identify suspicious transactions.