Breast carcinoma remains the most commonly diagnosed malignancy and a leading cause of cancer-related mortality among women worldwide. While mammography is the gold standard for early detection, challenges such as high breast density often obscure malignancies, reducing diagnostic sensitivity. Conventional parenchymal texture analysis methods have limitations due to struggles with spatial interpretation and noise sensitivity. This study investigates the efficacy of fractal-based global texture features for distinguishing between malignant and normal mammograms and assessing their potential for molecular subtype differentiation. Digital mammograms were analyzed using standardized preprocessing techniques, and fractal measures were computed to capture complexity and connectivity properties within breast tissue structures. Fractal dimension, multifractality strength, and succolarity reservoir were found to effectively characterize specific features of mammographic texture; however, their incorporation into machine learning models yielded moderate discriminatory performance between categories. We introduced succolarity reservoir as a novel parameter accounting for tissues latent connectivity. In addition, while succolarity reservoir exhibits potential for differentiating Luminal B from other molecular subtypes, its overall discriminative power remains limited. This proof-of-concept study underscores the potential of fractal-based texture analysis as a non-invasive biomarker in breast carcinoma diagnosis.