Fixing memory leaks in software is crucial to improving system efficiency, reliability, and stability. Nevertheless, manually fixing memory leaks is not only time-consuming but also error-prone. Recent research has proposed program repair techniques to fix memory leaks automatically. However, a common limitation of prior techniques is that they cannot tackle memory leaks in binary code. Given the huge popularity of binary code, it is highly necessary to fix its hidden memory leaks. To this end, this paper presents BFix, the first-ever binary-level memory-leak fixing tool. BFix creates effective binary analysis techniques to overcome the unique technical challenges of automatically and safely fixing memory leaks in binaries. Our experiments show that BFix can achieve promising fixing effectiveness and efficiency, comparable to state-of-the-art source-level fixing tools. Meanwhile, it has a negligible impact on binary code size and performance.