The use of smartphones is increasing and the leakage of user personal information is frequently occurring. Taint analysis tools to prevent such leakage of personal information include Flowdroid, Amandroid, and Scandroid, but they analyze Java byte cod...
The use of smartphones is increasing and the leakage of user personal information is frequently occurring. Taint analysis tools to prevent such leakage of personal information include Flowdroid, Amandroid, and Scandroid, but they analyze Java byte code only, not the native libraries. Even a few methods including Gangjin's research and Argus-SAF, which can analyze the native libraries, may not report correct results depending on the NDK versions. To solve this problem, we converted the native libraries into the LLVM intermediate representations according to the NDK version, and the resulting intermediate representations are analyzed using the taint analysis method. In order to measure the performance improvement of the proposed analyzer, it has experimented with 15 malicious applications and 15 normal applications. Experimental results show that the proposed method is superior to other methods, specifically, it recorded the highest F1-score 60.6% than others, FlowDroid 22.2% and Argus-SAF 41.4%.