This is solved in my paper here using a generalized version of the techniques of Sapir. We also show manysome other properties of the embedding can be achieved while adding malnormality, e.g that the preservation ofembedding is a q-i and preserves the decidability of the Word problem.