CRC Centre for research on Risks and Crisis December 10 2014 Big Data and maritime surveillance Aldo NAPOLI aldo.napoli@mines-paristech.fr Airborne Maritime Surveillance - Le Castellet 1
Outline MINES ParisTech-CRC and ongoing research Big Data : Definition and Origin Data Mining-based Environment Geovisual Analytics-based Environment Data Mining for Bayesian networks conception 2
MINES ParisTech-CRC and ongoing research 3
Centre for research on Risks and Crisis CRC French engineering school (since 1783) Research centre (since 2008) ( 1998 (Team created in Centre for research on Risks and Crisis French research association (since 1967) Research & training Business creation Industrial partnerships Strong industrial contacts 35 people: 11 Researchers, 15 PhD Students, 2 Engineers + support Engineering Sciences, Management science, Psychology, Computer science, Geography, Geographical information Sciences, Law. Develop research, teaching activities, methods and tools. Contribute to strengthen organisations and territories against disturbances. 4
Components of a Maritime Surveillance System 5
Our first responses to research challenges Since 2009: Modeling maritime risks Design of automatic or visual methods for exploring data and discovering knowledge (Big Data) Spatial data analysis using data mining or GeoVA Risks modeling Formalization of risk behaviours using ontologies and Bayesian networks Simulation of risk scenarios Simulation of the evolution of threats and responses to apply using Bayesian network Design of new surveillance systems Integration of automatic and visual analysis functions. 6
Maritime Surveillance System Engineering FishEye: the MSS of CRC Functions of the system: Monitoring the maritime traffic, Visualization of information used for behaviour analysis of ships at sea, Automatic and visual analysis of risk behaviours. 7
Big Data: definition and origin 8
Definition Big data is like teenage sex: everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it (Dan Ariely) Large volumes of data that can not be accommodated with traditional relational DBMS 5 V (Doug Laney Gartner-, 2001) Volume, Variety, Velocity (Veracity, Value) Challenges: storage, modelling, visualization of data, extraction of knowledge, etc. 9
A lot of data for maritime surveillance 10
A lot of data for maritime surveillance Meteorology Surface currents Waves Bathymetry Shoreline Etc. 11
Origin Matthew Fontaine Maury (1806-1873): pathfinder of the seas (Navy officer, oceanographer, catographer, meteologist, astronomer, historian, geologist, author, ) 12
Origin Physical Geography The Sea (1855) 13
Origin Physical Geography The Sea (1855) 14
Origin A scientific navigation method: toward maritime routes 15
Origin Whales and the Northwest Passage 16
ShipMINE: a data mining environment dedicated to risk behaviour analysis of ships at sea (PhD of B. IDIRI, 2013) 17
Data Mining 18
Fontionalities DBCSAN Periodica Découverte de comportements Découverte de comportements périodiques périodiques de navigation DBCSAN Détection de zones Détection à risques de zones denses Découverte de de convois navigations de trajectoires parallèles Convoy: Convoy: CuTS* CuTS* Typologie des Fonctionnalités méthodes de fouille de données TraClass TraClass Classification des navires des trajectoires selon leur comportement Découverte de trajectoires routes communes communes de navigation Détection de de trajectoires et sous- sous-trajectoires aberrantes anormales des navires TraOD TraOD TraClus TraClus 19 19
ShipMINE 20
Risk area detection Algorithm DBSCAN MAIB data base (Great Bretain, 1991-2009) Eps = 20 km MinPts = 50 9 risk areas 21
Identification of suspicious ships Algorithm CONVOY Trajectory of a supply vessel followed by a suspicious ship 22
Identification of shipping routes Algorithm TraClus Trajectories of supply vessels Eps = 10 km MinLns = 2 2 shipping routes 23
Detection of abnormal trajectories Algorithm TraOD Trajectories of 30 tankers in the Mediterranean sea P = 0.95 D = 20 km 12 Abnormal trajectories 24
A Geovizual Analytics environment dedicated to risk behaviour analysis of ships at sea (PhD of G. VATIN, 2014) 25
Geovizual Analytics Mallé-Noyon, 2008 Riveiro, 2011 Willems, 2011 Andrienko & Andrienko, 2013 26
Methodology Environnement User Goal Proposition of GeoVA solutions Interface of selection Knowledge-based System Ontologies Rules Contexte Problématique Etat de l'art Proposition Résultats Conclusion 27
Prototype Interface d utilisation ontologie Technologie Java (OWL API) Utilisable / téléchargeable depuis un navigateur Copie / lecture / écriture de l ontologie en local par l application 28
Using Data Mining with Bayesian Networks against piracy (PhD of A. Boujla, 2014) 29
Bayesian Network A model that represents knowledge, makes possible the calculation of conditional probabilities and provides solutions The problem of the response planning against a threat to offshore oil fields integrates strong constraints: Coordination between the different available counter-attack devices on the field, Real-time gradation of the threat and the response adaptation depending on its increase Inherent uncertainty of threat parameters, Automatization of the whole process. Bouejla 2014 30
Method Construction of Bayesian networks Bayesian network Data mining learning Brainstorming learning Database of the International Maritime Organization (IMO) Expert knowledge from the maritime security domain 31
Construction of BN using data mining 32
Simulation of attack scenarios Example of scenario Bayesian Network 33
GeoVizualisation in the MSS 34
Thanks for your attention Contact: Aldo NAPOLI MINES ParisTech CRC Phone: +33 4 93 67 89 15 E-mail: aldo.napoli@mines-paristech.fr Web: www.crc.mines-paristech.fr 35