Binary obfuscation plays an essential role in evading malware analysis and tampering with reverse engineering. Some widely used code obfuscation techniques focus on evading syntax based detection, however semantic analysis techniques have been developed to thwart their evasion attempts. Recently some binary obfuscation techniques with potential of evading both statistical and semantic detections have been proposed, taking concealment into account but lacking efficiency or security strength. This study proposes a binary obfuscation technique for mobile apps based on LZW and Huffman encoding to offer the potential of evading both statistical and semantic detections while taking intensity and concealment into account. This technique constructs the required instruction encoding tables. On one hand, it scrambles the sequence of original instructions with encoding tables to improve the intensity and concealment. On the other hand, it reinforces intensity by separating the encoding tables encrypted by white-box AES from code segment, concealing the key and lookup algorithm, in order to evading attacks on keys. A prototype tool for this technique, called ObfusDroid, is put forward, and an evaluation on ObfusDroid is given from aspects of intensity, cost, compatibility and concealment to demonstrate its capability of evading statistical analysis.