Abstract: Rich thread-level parallelism of GPU has motivated co-running GPU kernels on a single GPU. However, when GPU kernels co-run, it is possible that one kernel can leverage buffer overflow to ...