In　this　paper,　we　consider　the　problem　of　jointly　designing　transmit　precoding　(TPC)　matrix　and　receive　filter　matrix　subject　to　both　secrecy　and　per-transmitter　power　constraints　in　the　multiple-input　multiple-output　(MIMO)　interference　channel,　where　K　legitimate　transmitter-receiver　pairs　communicate　in　the　presence　of　an　external　eavesdropper.　Explicitly,　we　jointly　design　the　TPC　and　receive　filter　matrices　based　on　the　minimum　total　mean-squared　error　(MSE)　criterion　under　a　given　and　feasible　information-theoretic　degrees　of　freedom.　More　specifically,　we　formulate　this　problem　by　minimizing　the　total　MSEs　of　the　signals　communicated　between　the　legitimate　transmitter-receiver　pairs,　while　ensuring　that　the　MSE　of　the　signals　decoded　by　the　eavesdropper　remains　higher　than　a　certain　threshold.　We　demonstrate　that　the　joint　design　of　the　TPC　and　receive　filter　matrices　subject　to　both　secrecy　and　transmit　power　constraints　can　be　accomplished　by　an　efficient　iterative　distributed　algorithm.　The　convergence　of　the　proposed　iterative　algorithm　is　characterized　as　well.　Furthermore,　the　performance　of　the　proposed　algorithm,　including　both　its　secrecy　rate　and　MSE,　is　characterized　with　the　aid　of　numerical　results.　We　demonstrate　that　the　proposed　algorithm　outperforms　the　traditional　interference　alignment　algorithm　in　terms　of　both　the　achievable　secrecy　rate　and　the　MSE.　As　a　benefit,　secure　communications　can　be　guaranteed　by　the　proposed　algorithm　for　the　MIMO　interference　channel　even　in　the　presence　of　a　sophisticated/strong　eavesdropper,　whose　number　of　antennas　is　much　higher　than　that　of　each　legitimate　transmitter　and　receiver.