0 views collected around this technical thread.
After upgrading Kuaishou's Android project from AGP 3.4.2 to 7.0.1, the APK size unexpectedly grew despite faster builds, leading to a deep investigation of Dex compression, R8 changes, and build tool versions, ultimately revealing a fastutil library mismatch that was resolved by forcing the correct version.
The article details how Android developers can protect APK dex files by progressively hardening code—using dynamic loading, hooking, instruction extraction, java‑to‑C++ conversion, and ultimately DEX‑VMP virtual machine encryption—while outlining implementation steps, custom opcodes, JNI integration, and addressing compatibility and performance trade‑offs.
The paper presents a DEX‑VMP scheme that encrypts Dalvik bytecode and executes it via a custom virtual machine and JNI bridge, merging the strengths of dynamic loading, hooking, instruction extraction, and java‑to‑C++ conversion while highlighting compatibility issues, performance overhead, and the need for selective protection of high‑value Android methods.
In the second part of Baidu’s APK‑size‑optimization series, the team uses the Titan‑Dex framework to strip unnecessary build‑type and system Dex annotations—such as generic signatures and class‑relationship metadata—while automatically generating a whitelist via custom Lint detectors, ultimately shrinking the final Dex by about 1.2 MB.
This article explains the fundamentals of Android pluginization, covering class loading, resource handling, Dex and OAT formats, security considerations, and presents a lightweight plugin framework used by 58.com to reduce app size and enable dynamic updates.
The article explains that Android TV plugins can crash on Android 9 when the ART runtime inlines methods across dex files after an upgrade, and proposes a build‑time fix that scans for vulnerable call sites and injects lightweight try‑catch code to disable inlining, eliminating the crash.
The article explains why reducing APK size improves download conversion and performance, then details a systematic, divide‑and‑conquer approach for Baidu App that trims Dex, resources, native libraries and ZIP overhead using tools such as R8, AndResGuard, ByteX, Booster and AGP, achieving a reduction from roughly 123 MB to 115 MB with up to 6 MB saved per batch and outlining plans to open‑source the utilities.
The article explains the Android Dex format, its compact bytecode advantages, how d8 generates Dex files, details each structural component—from header and ID tables to class data and bytecode—covers encoding schemes, endianness, and demonstrates inspection with the dexdump tool for deeper ART understanding.
The article explains how Baidu’s Dex line-number optimization replaces original line numbers with compact PC-based mappings, unifies debug_info_items, and uses a server-side mapping file to shrink DebugInfo by roughly 8% of the Dex size while keeping stack-trace line numbers fully recoverable.
The Baidu App team applied a systematic, sustainable approach—combining Dex, resource, and native‑library optimizations with tools such as R8, ProGuard, AndResGuard, and custom compression pipelines—to shrink the APK by several megabytes, improve download conversion, and establish reusable mechanisms for future builds.
This article introduces DexInjector, a tool for directly instrumenting Android APK dex files to insert logging, performance monitoring, and third‑party hooks without recompiling, explains the underlying technical research, design choices, implementation details, handling of method limits, string jumbo, obfuscation, class conflicts, and provides code examples and workflow diagrams.
The platform optimizes Android APK size by shrinking resources, compressing images, filtering native libraries, and applying selective byte‑code reductions, then continuously monitors package composition, trends, and business‑line usage, alerting teams to abnormal growth while preserving functionality for large‑scale apps.
This article provides a comprehensive overview of Android app hardening methods, detailing the evolution from first‑generation dex and so protection to advanced code virtualization, and outlines various unpacking (脱壳) techniques—including memory dump, dynamic debugging, hook‑based and custom ROM approaches—used to bypass these protections.
The article explains Android’s DEX compilation flow, the transformation of DEX into VDEX and ODEX during ART pre‑optimization, why ODEX lacks the original classes.dex causing ARouter routing failures, and how disabling dex stripping in the build restores the missing classes for stable navigation.
This article explains Android code obfuscation, compares ProGuard and R8, details their four core functions, shows how to enable or disable them via Gradle settings, discusses ProGuard rule syntax and sources, and provides best practices for modular projects and componentized obfuscation.
This article explains why Android hot‑fixes are needed, describes the structure and generation of class and dex files, compares popular hot‑fix frameworks such as Tinker and AndFix, and provides practical guidance on integrating and managing hot‑fix solutions in mobile apps.
This article explains how to use Python to analyze Android APK packages, extract basic statistics, identify optimizable resources such as oversized images, duplicate files, and unused assets, and provide data‑driven guidance for reducing APK size and improving distribution efficiency.
This article provides an in‑depth examination of Android's ClassLoader architecture, explaining how BaseDexClassLoader.findClass() and ClassLoader.loadClass() locate and load classes, with step‑by‑step source code analysis of loadClass, findLoadedClass, findClass, DexPathList, DexFile, and native class definition processes.
The article explains what a Dex file is, how it is produced from Java .class files via the dx tool after JVM compilation, and why understanding Dex is crucial for Android tasks such as APK slimming, hot‑fixes, plugin architecture, app hardening, reverse engineering and overcoming the 64K method limit.
This article explains the origin of Android's 64K method limit, how it varies across OS versions, and evaluates several approaches—including Google Multidex, asynchronous child‑process loading, plugin architectures, and lazy loading—to help developers choose the most suitable solution for their app.