On Reading SRAMs in IR Images, and Establishing Bounds on Trust