A Context-Sensitive Framework for Humanitarian Data Protection: Integrating Privacy, Biometrics, Surveillance, and AI Risks

Digitalization has become integral to humanitarian registration, communication, assistance delivery, coordination, and protection. The same infrastructures can, however, increase the scale and persistence of harm when personal data are collected excessively, retained unnecessarily, shared opaquely, processed biometrically, or exposed to AI-enabled information manipulation. This study develops a context-sensitive humanitarian data-protection framework that integrates data governance, privacy engineering, biometric and surveillance safeguards, and AI/information-integrity governance. The empirical component uses a bilingual Arabic/English scenario-based questionnaire administered online. The supplied dataset contains 130 records; 128 respondents provided consent and were retained for the primary analysis. The study combines descriptive statistics, internal-consistency diagnostics, non-parametric comparisons, exploratory subgroup analysis, safeguard prioritization, and a preliminary exploratory factor analysis of the 15 risk-domain items. The three principal risk domains showed exploratory internal consistency (Cronbach's α = .711–.737). Mean perceived risk was highest for privacy and data governance (M = 4.063, SD = 0.744), followed by biometric/surveillance risk (M = 3.834, SD = 0.770) and AI/misinformation risk (M = 3.742, SD = 0.798). A Friedman test indicated a significant within-respondent difference across the three domains, χ²(2) = 29.49, p < .001. Incident response and notification, transparent data sharing, and data minimization were among the most frequently selected safeguards. The preliminary EFA was based on the 15 C-E risk items and used the available pairwise correlation matrix; sampling adequacy was strong (KMO = .882) and Bartlett's test of sphericity was significant, χ²(105) = 566.57, p < .001. Eigenvalues supported a parsimonious three-factor solution, although several items showed cross-loadings; the EFA is therefore treated as preliminary structural evidence rather than definitive validation. Because the sample was non-probabilistic and was not drawn directly from a defined refugee or humanitarian-worker population, findings are interpreted as exploratory perceptions rather than population estimates. The contribution is consequently a transparent, empirically informed framework-development model with a clear pathway toward expert content validation and larger-sample confirmatory testing.