Aperçu du cours
This pioneering course blends the domains of cyber security and artificial intelligence. It has been designed for cyber security professionals who want to understand and implement AI models for exploring logs, security events, and other types of data. Classifying analytics is especially encouraged as well as NLP and Image Recognition techniques based on types of data sources found in the cybersecurity domain.
Where coding is needed, Python and some selected libraries will be used. The expected audience is expected to be familiar with scripting coding but is not required to master any specific language.
Prérequis
- Python or similar scripting language (like R, Matlab, etc)
- Notions in AI/Machine Learning
Objectifs pédagogiques
- Understand the concepts of AI, ML and DL. Their strengths and limitations
- Generate different visualisations of your data by applying statistical models to real cybersecurity problems in meaningful ways
- Familiarise with ML frameworks and methods
- Identify the best suited ML models to solve complex problems
- Work in specific cybersecurity use cases while being supervised by an AI expert
Public ciblé
- Engineers
- Developers
Programme de formation
-
Introduction to AI, ML & DL
-
DL as an approach to AI
-
Neural Networks and Types
-
Data Types
-
Strengths and Limits of ML
-
-
Sample Toy Study
-
Image recognition
-
Complex regressions Use Case
-
-
AI Frameworks
-
Keras as Reference & Doc Framework
-
-
Framework Setup & Workshop
-
PyTorch installation and Setup
-
Toy-study in Pytorch
-
-
Introduction to NLP
-
NLP and Applications
-
Data Cleaning & Preprocessing
-
Tokenization
-
Stop-words, stemming & lemmatization
-
Text Data Vectorization
-
BERT, Transformers (and Adapters)
-
-
Classification & Clustering
-
Theory
-
KNN, K-Means
-
-
NLP Sample Toy Study
-
Text classification & clustering (using scikit-learn)
-
-
Interactive Discussion
-
Put the toy-study results into test
-
-
Additional Supervised & Unsupervised methods
-
Supplemental Methods to be defined
-
Model selection and evaluation
-
Visualisation
-
-
AI in Cybersecurity
-
Analytics based on data sources
-
AI Topics in Cybersecurity
-
-
NLP Applied to Cybersecurity
-
Introduction
-
Network threat analysis
-
Text Classification Methods to Detect Malware
-
Process Behaviour Analysis
-
Abnormal system behaviour detection
-
-
Use Case 1
-
Using ML to Detect Malicious URLs
-
(Note: This can be replaced by a Transformers use-case: using a pre trained BERT model, i.e. from hugging face)
-
-
Interactive Discussion
-
Performance discussion, visualisation and comparison
-
-
Statistical Methods in ML
-
Intuition vs Statistics
-
Univariate Numerical Analysis
-
Bivariate Numerical Analysis
-
-
Examples
-
Univariate (Mean, Median, Percentile, SD)
-
Bivariate (Correlation, Pearson Correlation)
-
-
More Statistics
-
Skewness & bias estimations
-
(basic) Bayes & Max entropy methods
-
Confidence Level approach
-
-
Case 1 (Continuation)
-
Statistical Methods as a tool for performance hypothesis testing
-
Improve the DL model’s performance
-
-
AI in Cybersecurity (part 2)
-
Transaction fraud detection
-
Text-based malicious intent detection
-
Machine vs human differentiation – or – Business data risk classification
-
-
Use Case 2
-
Using ML as Intrusion Detection System
-
ML for Same Person Identification (prefered choice)
-
-
Revision & Closing
-
Interactive Multiple Choices Questions Revision
-
Closing Words
-
Instructeur
