Recently, metal-oxide based resistive-RAM (RRAM) has emerged as a promising nonvolatile memory (NVM) candidate. However, it is also recognized that the stochastic RRAM switching mechanisms inevitability introduce large variability in device parameters, which impose severe challenges for memory applications especially in high-density arrays. RRAM variability can be attributed to the microscopic variation of dimension and/or composition of the filamentary conductive paths during the switching process. Variability in switching voltage/current reduces safe operation margin and resistance variation degrades sensing margin. Cell-to-cell and cycle-to-cycle variability has different origins and needs to be differentiated in RRAM characterization. Resistance variability measured by /µ (deviation/mean) has to be controlled within target ranges determined by device and array specifications. Mechanism and typical characteristics of RRAM variability will be reviewed in this presentation. Although variability is undesirable for memory arrays, security applications embrace truly random variations. Pervasive and ubiquitous computing requires robust light-weight security technologies at low cost. Physical unclonable functions (PUF) exploit physical randomness and variability as security primitives. RRAM variability may be utilized for PUF implementation, which may achieve much smaller footprint than existing PUF solutions. Feasibility of PUF designs based on RRAM variability will be analyzed based on measured device properties