The unit normal to a two dimensional surface in 3 space maps the surface into the unit sphere. The induced vector bundle under this map, the Gauss map,is the surface's tangent bundle as can be seen from parallel translating its tangent planes to the origin.
The induced connection on the surface's tangent bundle - i.e. the pull back under the Gauss map of the standard connection on the sphere - defines a connection on the surface.
Here are some questions about the induced connection.
Given a second embedding of the surface, the Gauss map induces a second connection. This second Gauss map is homotopic to the first - since both classify the surface's tangent bundle. One can imagine deforming one surface into the other to obtain a homotopy and simultaneously a homotopy between the two connections. Does such a deformation always exist? If not can the homotopy be realized as a 1 parameter family of embeddings in a higher dimensional Euclidean space?
Is the induced connection just the Levi-Civita connection on the surface that it inherits from its embedding in 3 space?
If so this seems quite strange since the derivative of the Gauss map is not usually the induced map of vector bundles.