<p>Age-related changes in facial appearance pose a significant challenge to reliable face recognition in longitudinal and real-world applications, often degrading the performance of appearance-based and deep learning approaches. This study addresses the problem of age-invariant face recognition by proposing a novel set of graph-based facial biomarkers that capture stable geometric relationships among facial landmarks. Our approach extracts a compact 5-dimensional feature vector from facial landmark graphs using five spectral energy measures: Graph Energy (GE), Laplacian Energy (LE), Domination Energy (DE), Distance Energy (DistE), and Signless Laplacian Energy (SLE). To assess the age-invariance and discriminative strength of these biomarkers, an extensive pairwise separation analysis was performed across 3,321 unique subject pairs drawn from a dataset of 82 individuals. The results demonstrate that all five features are highly effective in distinguishing individuals, as evidenced by strong mean separation scores. Furthermore, a one-sample t-test confirmed the statistical significance of each feature, reinforcing their validity as robust age-invariant biomarkers. The proposed framework establishes a learning-free, interpretable, and computationally efficient baseline for longitudinal facial analysis and age-invariant face recognition. Future work will focus on improving robustness under real-world conditions by investigating alternative graph constructions, feature aggregation across multiple images, and integration with supervised learning models for large-scale recognition tasks.</p>