Central venous catheters are often filled when not in use with an anticoagulating fluid, usually heparinized saline, known as the locking fluid. However, the use of the locking fluid is associated with known risks because of "leakage" of the lock. A new hypothesis is proposed here to explain the lock fluid leakage: that the leakage is due to advective and diffusive mass transfer by blood flow around the catheter tip in situ. On the basis of previous in vitro experiments, the leakage mechanism has been hypothesized to be fluid motion driven by buoyancy forces between the heavier blood and the lighter locking fluid. The current hypothesis is justified by a simple one-dimensional mass transfer model and more sophisticated three-dimensional computational hemodynamic simulations of an idealized catheter. The results predict an initial, fast (<10 seconds) advection-dominated phase, which may deplete up to 10% of the initial lock, followed by a slow diffusion-limited phase which predicts an additional 1-2% of leakage during a 48 hour period. The current results predict leakage rates that are more consistent with published in vivo data when compared with the buoyancy hypothesis predictions, which tend to grossly overestimate leakage rates.