With the rapid growth of OpenHarmony, a new distributed operating system (OS), security and privacy issues have become major concerns. While taint analysis has proven effective in Android, OpenHarmony still lacks a comparable framework. Additionally, OpenHarmony's programming language, ArkTS, has several unique features compared to other languages like TypeScript and Java, including complex lifecycles, closure mechanisms, and a distinct API ecosystem. Directly adapting existing taint analysis tools for Android cannot achieve effectiveness in OpenHarmony applications. To address this challenge, we propose HapFlow , a novel taint analysis framework tailored specifically for the OpenHarmony platform and ArkTS programs. This work presents: (1) an LLM-assisted method for identifying and categorizing source APIs for OpenHarmony, (2) a lifecycle and callback functions modeling approach that supports declarative UI in ArkTS, and (3) an IFDS-based taint propagation extension that accurately handles closure-induced cross-function data flows in OpenHarmony applications. Therefore, HapFlow enables precise inter-procedural data-flow analysis for OpenHarmony applications. We validate HapFlow 's effectiveness on the HapBench benchmark, achieving 96.15% precision and 94.34% recall. Furthermore, HapFlow identifies 73 sensitive data leak flows across over 3,000 open-source projects with 8 false positives, completing over 98% of analyses within 10 seconds. These results demonstrate HapFlow 's practicability and scalability for taint analysis in OpenHarmony applications.
Chen et al. (Tue,) studied this question.