Repackaged App Detection via Label Sound and Appearance Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting repackaged mobile apps, particularly in Android operating systems, face challenges due to high rates of false positives and false negatives when using exact string matching, especially with non-English app labels, which can be difficult to distinguish visually or phonetically.
Innovation Solution
A system that compares the similarity of app labels between a target and a reference mobile app based on sound and appearance, using phonetic algorithms like Soundex for English and Pinyin/Romaji for non-English languages, along with image comparison, while also verifying identifiers to determine if a mobile app has been repackaged.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If exact string matching is used to detect repackaged mobile apps, then detection precision may be improved for identical app labels, but false positive and false negative rates increase when app labels have subtle variations
Solution Approach 1:
The patent transforms the detection approach from exact string matching to similarity-based comparison by changing the parameter from binary equality to continuous similarity scores. This involves converting app labels into phonetic representations (Soundex, Pinyin, Romaji) and visual features, then computing similarity metrics that allow for subtle variations while maintaining detection accuracy.
Solution Approach 2:
The patent replaces the mechanical exact string matching system with a multi-dimensional similarity comparison system. Instead of direct character-by-character comparison, it substitutes phonetic algorithms that capture sound similarities and image processing that captures visual similarities, thereby resolving the contradiction between precision and reliability.
2Reliability
If phonetic algorithms and image comparison are used to compare app labels, then detection reliability is improved by reducing false positives and negatives, but device complexity increases
Solution Approach 1:
The patent segments the app label comparison task into distinct modules: phonetic conversion module, visual feature extraction module, similarity computation module, and identifier verification module. Each module handles a specific aspect of the comparison, making the complex system more manageable and maintainable while improving detection reliability through comprehensive multi-dimensional analysis.
Solution Approach 2:
The patent creates a universal detection framework that handles multiple languages and scripts through phonetic algorithms (Soundex for English, Pinyin for Chinese, Romaji for Japanese) and a unified image comparison approach. This multi-functional system can process diverse app labels across different languages using the same underlying methodology, reducing overall system complexity despite the enhanced capabilities.
3Ease of operation
If app labels are compared based on sound and appearance similarity, then ease of operation is improved by handling non-English languages, but measurement precision may be reduced for visually similar but phonetically different labels
Solution Approach 1:
The patent adds multiple dimensions to the comparison process: phonetic dimension (sound similarity), visual dimension (appearance similarity), and identifier dimension (technical uniqueness). By evaluating app labels across these multiple dimensions rather than relying on a single dimension, the system achieves better overall accuracy in distinguishing repackaged apps while maintaining ease of operation across different languages.
Solution Approach 2:
The system incorporates feedback mechanisms where the results from phonetic comparison, visual comparison, and identifier verification are combined to reach a final determination. This feedback loop allows the system to cross-validate findings across different comparison methods, improving measurement precision by identifying cases where one dimension may be ambiguous but another provides clear differentiation.
Data Source
AI summary
A repackaged mobile app that has been unpacked and repackaged back is detected based on similarity of app labels of a target mobile app being evaluated and a reference mobile app. The similarity of the sound of the app label of the target mobile app to the sound of the app label of the reference mobile app may be determined. The similarity of the appearance of the app label of the target mobile app to the appearance of the app label of the reference mobile app may also be determined. The target mobile app may be deemed to be a repackaged mobile app when the app labels of the target and reference mobile apps are deemed to be similar (which may include being the same) but the target and reference mobile apps have different identifiers.