We introduce a lightweight, cost-effective, and portable solution for real-time peripheral oxygen saturation (SpO2) measurement using photometric sensing through a smartphone's camera. Specifically, we design a hardware plug-in module that snaps onto the smartphone's flashlight and estimates the blood oxygen content from the light intensity reflected off the user's finger and registered on camera images. The oxygen levels are mapped to equivalent photoplethysmography (PPG) signals used for the SpO2 estimation using a machine learning based one-time calibration. With the knowledge that blood oxygen largely responds to Infrared (IR) and Red wavelengths, state-of-the-art pulse oximetry techniques use IR and RED light emitting diodes and photodetectors to sense each channel. We further develop a novel solution that exploits the IR leakage of the LED white light of the smartphone. The system is incorporated with a hardware of IR and RED filters that are spatially separated such that the respective signals are registered on independent areas of the image sensor. We present the preliminary results and analyse possible challenges for further improvement.