Women life suffered and killed by invasive cancerous tumors is the most frequently highlighted header in many of the newsletters since 2010. There can be many. By their nature invasive tumors spread from tissue to tissue, abduct themselves to cause new tumors. In many of the human biological parts, CT scan has given an accurate approach for early and successful detection of the cancerous tumors. But there were cases where diagnosis with CT scan images failed and resulted in many false positives. In economically backward countries these false positives raised the notifying concern from women who cannot afford multiple diagnosis tests. Due to change in biological metabolism the growth of the breast fat in women to considerably abnormal size is the main cause for false positives. In many of the images under study, this huge thick breast fat layer led to the rise of misclassification rate with false positives. Rendering societal help is a need for an accurate mechanism that can reduce the false positives at an initial diagnosis itself. The proposed method introduces a novel constraint-based algorithm to classify a mammogram image as cancerous, aiming to reduce false positives. Our proposed deep learning algorithm WOMT is trained with Wasserstein Distribution constraints that are derived from mass transfer of cancerous patches to non-cancerous patches. Our experimental simulations with a deep learning model trained with these constraints resulted in reduced false positives.
Wearable sensors are increasingly deployed in healthcare, fitness, and human activity recognition (HAR). Adversarial attacks on deep learning models used in HRA increase security concerns in consumer goods. Adversarial time-series perturbations decrease the accuracy of prediction models. This paper investigates the impact of Time-Warping Adversarial Attacks (TWAA), which manipulate the temporal axis of sensor signals while preserving amplitude ranges. Thereby, generating adversarially misleading samples. We propose a trust-centric framework comprising attack execution, detection, and mitigation modules, and evaluate it on two benchmark datasets (UCI HAR, WISDM and MHEALTH) using baseline deep models (CNN, LSTM, and CNN–LSTM). Experimental results show that TWAA significantly degrades classification performance. It reduces the adversarial accuracy by up to 30% and achieving attack success rates exceeding 40% at moderate perturbation parameters. Our detection mechanism achieves AUROC > 0.85 across different types of TWAA, while the mitigation strategy restores model accuracy to above 80% with inference latency under 0.01 ms/sample. These findings demonstrate both the severity of temporal adversarial threats in HAR and the effectiveness of our proposed trust-centric defense. Unlike existing mechanisms that rely on adversarial training or offline robustness enhancement, the proposed framework enables runtime detection of physically plausible time-warping adversarial attacks and applies conditional input-level temporal repair using dynamic time warping. Thus, providing a holistic pipeline for the analysis, detection and mitigation of TWAA on wearable consumer goods.