There are different ways to approach it, as a purely 2d composite, some kind of 2.5D mix or 3D.
For 2D way, track the eyes (I’d suggest track pupil because it is the most prominent feature and easiest to follow) and then matchmove and warp the cataract effect on top of eye surface. For deformations use warps: gridwarp, spline based warp, whichever feels comfortable.
For 2.5D way, do a 3d track of head using some suitable tool, deduce eye locations in screen space from that track and then follow up as in 2D way.
For full 3D, do head matchmove, animate and render 3D eyes and comp on top. The hard part here is probably animating eye rotations, it will be pretty tricky to get a 3d solve for eyes automatically, I’d rather hope for good head track and then hand-animate eyes on top of that.
How big are the eyes in shot? Is it a closeup or wide shot? If eyes are small, you will have more trouble tracking them, but as an upside, any slippage is also less visible. Closeups are easier to track but also make mistakes very easy to see. If your cataract starts to slide on eye surface you loose the credibility of effect.
If you are a beginner, I’d probably suggest the easy hard way: manually animate your eyes over your video in 3D scene. It will teach you a lot about animation, why automatic tracking is useful, why it fails (if you can’t follow something by eye (pun intended), how on earth shoud autotracker find it?) and so on.