The Efficiency of Learning Methodology for Privacy Protection in Context-aware Environment during the COVID-19...

Area of Science:

• Computer Science
• Information Security
• Machine Learning

Background:

• The COVID-19 pandemic led to widespread lockdowns, altering user behavior and privacy expectations.
• Context-aware applications rely on user behavior patterns, which were significantly disrupted during the pandemic.
• Machine learning (ML) algorithms trained on pre-pandemic data may experience performance degradation due to these behavioral shifts.

Purpose of the Study:

• To assess the impact of the COVID-19 pandemic on the performance of a privacy protection system.
• To evaluate the efficiency of a hybrid ML methodology within the Privacy Preferences Manager (PPM) module under pandemic conditions.
• To quantify the changes in accuracy and F1 score of the ML algorithm.

Main Methods:

• The study focused on the Privacy Preferences Manager (PPM) module of a privacy protection system.
• A hybrid methodology combining a Statistical Model (SM) and Logistic Regression (LR) was employed.
• The hybrid model's efficiency was evaluated using two real-world datasets collected before and during the COVID-19 pandemic.

Main Results:

• The COVID-19 pandemic significantly impacted the efficiency of the hybrid ML methodology.
• Accuracy of the PPM module decreased by 13.05% due to pandemic-induced behavioral changes.
• The F1 score of the PPM module decreased by 15.22% during the pandemic period.

Conclusions:

• Pandemic-related shifts in user privacy preferences negatively affect the performance of ML-based privacy systems.
• The hybrid SM-LR methodology's efficiency is demonstrably reduced under the extraordinary circumstances of a global pandemic.
• Future privacy protection systems may require adaptive learning mechanisms to account for significant behavioral disruptions.