PRISM: Refracting Digital Influence into Ethical, Adaptive Advertising A Privacy-preserving, Real-time, Interoperable, Structure-aware and managerially governed framework for Social Network Analytics in e-commerce advertising

Abstract Social Network Analytics (SNA) promises influence-driven advertising, yet deployed systems still rank influencers once, on one platform, and treat privacy law as an afterthought. We propose PRISM, a five-layer framework that addresses these problems together. PRISM (i) computes influence only from noised node-level scores released under ε-differential privacy, aligned with the GDPR and India's DPDP Act 2023; (ii) keeps influence maps current through exponentially decayed temporal centrality; (iii) fuses platforms into a multiplex graph linked by a pseudonymous key rather than raw profiles; (iv) allocates seeds with awareness of community structure; and (v) embeds the analytics in a cross-functional governance loop. We formalise PRISM, state three propositions, and test them in a reproducible simulation of a two-platform market (2,000 users, 6 periods, 15% edge rewiring per period, Independent Cascade diffusion, 5 replicates). PRISM reached 9.2% more adopters than static single-platform degree seeding on average, and 16% more by the final period. With ε = 1 it still beat both single-platform baselines and overtook every static baseline from period 5. Reach stayed within one standard deviation of the non-private level down to ε = 2. An ablation showed that community quotas slightly reduced reach, so diversity quotas are better justified by brand-safety goals than by reach. We close with managerial, regulatory and research implications.