This paper presents a novel approach and algorithm to the problem of magnetic field interference cancellation of time-varying interference using distributed magnetometers and spacecraft telemetry with particular emphasis on the constrained computational and power requirements of CubeSats. The traditional approach to enable space-based low-amplitude and low-noise magnetometry is to develop a spacecraft magnetic cleanliness design and place the magnetometer sensor at the end of a boom far enough away from the bus to minimize remaining stray magnetic fields. In addition, secondary magnetometers are often placed partway along the boom to apply magnetic field gradiometry to clean the data further (e.g., NASA MMS has 8 meter booms with a sensor half-way down and another at the end). We employ a different approach taking advantage of low-cost chip-based magnetometers that can be placed throughout the satellite bus instead of utilizing a boom. The spacecraft magnetic field interference cancellation problem that we solve involves estimation of noise when the number of interfering sources far exceeds the number of sensors required to decouple the noise from the signal. The proposed approach models this as a contextual bandit learning problem and the proposed algorithm learns to identify the optimal low-noise combination of distributed magnetometers based on indirect information gained on spacecraft currents through telemetry. The algorithmic behaviors are tested with synthetically modeled spacecraft data and on real world data generated in a lab-based setting with telemetry and currents collected from the GRIFEX CubeSat and provides a way for accurate magnetic field measurements with CubeSats.