Smart Grid Security: An Effective Hybrid CNN-Based Approach for Detecting Energy Theft Using Consumption Patterns.

Area of Science:

• Computer Science
• Electrical Engineering
• Cybersecurity

Background:

• Smart grids rely on smart meters for data collection, crucial for load monitoring and energy management.
• Energy theft via data tampering by malicious consumers poses a significant global challenge, causing substantial financial and technical losses.
• Existing methods struggle to effectively detect sophisticated data manipulation attacks on smart meter readings.

Purpose of the Study:

• To develop and evaluate a hybrid deep learning system for detecting energy theft in smart grids.
• To address the challenge of data tampering cyber-attacks targeting smart meter consumption data.
• To improve the accuracy and robustness of energy theft detection systems.

Main Methods:

• A hybrid system combining Convolutional Neural Networks (CNN) for feature extraction and traditional machine learning algorithms for classification was developed.
• Six distinct data tampering attack vectors were simulated to generate realistic tampered datasets.
• The Generative Adversarial Network (GAN) method was employed to address data imbalance and enhance model training with synthetic data.
• Both specialized detectors for individual attack vectors and a general detector for all attacks were designed and evaluated.

Main Results:

• The proposed hybrid CNN-based system demonstrated satisfactory accuracy in classifying both honest and malicious users.
• The general detector, trained on a dataset encompassing all attack vectors, proved effective in identifying energy theft.
• The application of GAN for data augmentation significantly improved the training process for the detection system.

Conclusions:

• The developed hybrid CNN-based system offers a promising solution for detecting energy theft in smart grids.
• The proposed approach effectively identifies data tampering cyber-attacks, mitigating non-technical losses.
• This research highlights the potential of deep learning and GANs in securing smart grid infrastructure.