What are the steps needed...
What are the steps needed...
- Feature Extraction
- Data Preparation
- Artificial Intelligence
- Concept Drift
Feature Extraction
Feature Extraction
- Static, Dynamic, Memory features
- This is not the main goal of work
- The idea in this step is to extract some information from PE file in order to classify it
Data Preparation
Data Preparation
- Create an input that compatible to ML model
- Delete duplications
- Delete redundant features
AI
AI
This is where concept drift analysis begins
AI
- Train a ML model using initial period data
- Test it with subsequent data
- Calculate selected performance metrics
- F1 Score
- FPR
Concept Drift
Concept Drift
- Concept Drift Detection
- Concept Drift Understanding
- Concept Drift Adaptation
Concept Drift Detection
- Observing performance changes in dataset from different timeline
- Classify as:
- Sudden Drift
- Incremental Drift
- Gradual Drift
- Reoccurring Drift
Concept Drift Understanding
- Where? When? How?
- What triggered to drift
- API deprecations, API advancements etc...
Concept Drift Adaptation
Update the model
- Online Learning
- Ensemble Methods
What have I done so far.
Timestamps
Timestamps
Feature Extraction
Feature Extraction
Data Preparation
Data Preparation
AI
AI
AI
Concept Drift
Concept Drift
Conclusion
Static Feature
Memory Feature
Merged Feature
Drift Classification
Future Work
Future Work
- More Analysis (sample) for the reliable results
- Using memory dump effectively
- Investigate reasons for the drift
- Strengthen the model