Visual Data Analysis
in cybersecurity education and forensic anthropology
Radek Ošlejšek – Faculty of Informatics, MU
Cybersecurity education
●
Cybersecurity (training)
– Software: KYPO Cyber Range Platform, KYPO Analyst
– In cooperation with CERIT (Pavel Čeleda, Jan Vykopal, Jakub Čegan, …)
PV226 Introductory seminar2
Cybersecurity: KYPO Analyst
PV226 Introductory seminar3
Events log
User1;2.08.2020 10:31:43;use webmin_backdoor
User1;2.08.2020 10:32:44;set RHOST
User1;2.08.2020 10:33:19;set LHOST
User1;2.08.2020 10:34:27;set SSL
User1;2.08.2020 10:34:35;set TARGET
User2;2.08.2020 10:32:17;use webmin_backdoor
User2;2.08.2020 10:32:43;exploit
User2;2.08.2020 10:44:33;set RPORT
User2;2.08.2020 10:45:21;exploit
User2;2.08.2020 10:56:02;set LHOST
User2;2.08.2020 10:56:20;set SSL
User2;2.08.2020 10:58:35;set TARGET
●
Goal: Support post-training analysis (revealing flaws in training design, difficulty, gameplay
strategies, etc.)
●
Techniques: process mining, metric-based data analysis, clustering and other ML
methods.
Input: Process-oriented data
Examples of Bachelor and Master Thesis
●
Integration and optimization of process graphs
– Currently only Heuristic net is available that suffers from inefficiency
PV226 Introductory seminar4
Forensic Anthropology
●
3D Face Identification and Forensic Analysis
– Software: FIDENTIS Analyst II
– In the cooperation with Department of Anthropology, Faculty of Science (Petra Urbanová)
PV226 Introductory seminar5
Forensic Anthropology
PV226 Introductory seminar6
Input: Photogrametry data
(3D geometry + photo texture )
●
Automated face identification in big data sets
(100.000+ faces).
●
Pre-selection of (a few) thousands of candidates.
●
Fully automated process.
●
Forensic-aware (anatomically correct) identification on
smaller data sets.
●
Automated pre-processing with exploratory analysis
and expert-driven decision-making.
●
Visual-analysis methods of detail forensic
investigation.
Goals:
3D Face Similarity Measurement
●
Goals: To decide, whether the two faces belong to the same person and why.
– Computation/enumeration of (dis)similarity of 3D scans
– Providing a visual representation for decision-making of forensic analysts.
PV226 Introductory seminar7
Descriptor-based Similarity Techniques
●
Descriptors: Color of eyes, distance between eyes, curvature of some are (e.g., nose), etc.
●
Descriptors must be detected automatically.
●
These techniques are independent on the position of 3D scans in space.
PV226 Introductory seminar8
Registration-based Similarity Techniques
●
Automated registration (mutual alignment in space) followed by space-dependent measurement
techniques
PV226 Introductory seminar9
Examples of Bachelor and Master Thesis
●
Automated Anthropological Landmark Detection (using machine-learning techniques)
PV226 Introductory seminar10
Examples of Bachelor and Master Thesis
●
Automated pose detection (orientation of facial scans in the space)
– Geometrically and/or using machine-learning methods
PV226 Introductory seminar11
Examples of Bachelor and Master Thesis
●
Superimposition of 3D face scan to body scan
PV226 Introductory seminar12
PhD Topics
Even the scans of the same person are not the same. We need identification techniques that can adapt
to changes introduced by
PV226 Introductory seminar13
PhD Topics
Even the scans of the same person are not the same. We need identification techniques that can adapt
to changes introduced by
●
Ageing: identification and simulation/prediction
PV226 Introductory seminar14
[Urbanová et. al, 2020]
PhD Topics
Even the scans of the same person are not the same. We need identification techniques that can adapt
to changes introduced by
●
Ageing: identification and simulation/prediction
●
Poses
PV226 Introductory seminar15
[Zhou, S., Xiao, S.: 3D face recognition: a survey. 2018]
PhD Topics
Even the scans of the same person are not the same. We need identification techniques that can adapt
to changes introduced by
●
Ageing: identification and simulation/prediction
●
Poses
●
Occlusion
●
...
PV226 Introductory seminar16
[Zhou, S., Xiao, S.: 3D face recognition: a survey. 2018]
PhD Topics
Even the scans of the same person are not the same. We need identification techniques that can adapt
to changes introduced by
●
Ageing: identification and simulation/prediction
●
Poses
●
Occlusion
●
Variability in a given population (i.e., computationally demanding statistical exploration and
evaluation)
●
...
PV226 Introductory seminar17
Thank you for your attention!
PV226 Introductory seminar18
●
KYPO Analyst: www.radek-oslejsek.cz/it/cybersecurity-education-and-training
●
FIDENTIS Analyst II: www.radek-oslejsek.cz/it/fidentis-analyst-2
●
oslejsek@fi.muni.cz