Entropy-Based Non-Invasive Reliability Monitoring of Convolutional Neural Networks

Convolutional Neural Networks (CNNs) have become the foundation of modern computer vision, achieving unprecedented accuracy across diverse image recognition tasks. While these networks excel on in-distribution data, they remain vulnerable to adversarial perturbations imperceptible input modifications that cause misclassification with high confidence. However, existing detection methods either require expensive retraining, modify network architecture, or degrade performance on clean inputs. Here we show that adversarial perturbations create immediate, detectable entropy signatures in CNN activations that can be monitored without any model modification. Using parallel entropy monitoring on VGG-16, we demonstrate that adversarial inputs consistently shift activation entropy by 7% in early convolutional layers, enabling 90% detection accuracy with false positives and false negative rates below 20%. The complete separation between clean and adversarial entropy distributions reveals that CNNs inherently encode distribution shifts in their activation patterns. This work establishes that CNN reliability can be assessed through activation entropy alone, enabling practical deployment of self-diagnostic vision systems that detect adversarial inputs in real-time without compromising original model performance.

LSTM-based Load Forecasting Robustness Against Noise Injection Attack in Microgrid

In this paper, we investigate the robustness of an LSTM neural network against noise injection attacks for electric load forecasting in an ideal microgrid. The performance of the LSTM model is investigated under a black-box Gaussian noise attack with different SNRs. It is assumed that attackers have just access to the input data of the LSTM model. The results show that the noise attack affects the performance of the LSTM model. The load prediction means absolute error (MAE) is 0.047 MW for a healthy prediction, while this value increases up to 0.097 MW for a Gaussian noise insertion with SNR= 6 dB. To robustify the LSTM model against noise attack, a low-pass filter with optimal cut-off frequency is applied at the model's input to remove the noise attack. The filter performs better in case of noise with lower SNR and is less promising for small noises.