JSidentify-V2: Leveraging Dynamic Memory Fingerprinting for Mini-Game Plagiarism Detection
Zhihao Li 2 4 , Chaozheng Wang 2 3 4 , Zongjie Li 5 6 , Xinyong Peng 2 , Qun Xia 2 , Haochuan Lu 2 , Ting Xiong 2 , Shuzheng Gao 3 , Cuiyun Gao 3 , Shuai Wang 6 , Yuetang Deng 2 , Huafeng Ma 2 · 2025
Read JSidentify-V2: Leveraging Dynamic Memory Fingerprinting for Mini-Game Plagiarism Detection by Zhihao Li 2 4 , Chaozheng Wang 2 3 4 , Zongjie Li 5 6 , Xinyong Peng 2 , Qun Xia 2 , Haochuan Lu 2 , Ting Xiong 2 , Shuzheng Gao 3 , Cuiyun Gao 3 , Shuai Wang 6 , Yuetang Deng 2 , Huafeng Ma 2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-01655.txt.
Introduction
Zhihao Li and eleven co-authors catch copied mini-games by fingerprinting how their code behaves in memory while it runs. Their 2025 paper faces a platform problem: popular small games are stolen, re-skinned, and republished. Static code scans catch lazy copies, but deep obfuscation, such as code encrypted with remote keys, destroys the code's shape and blinds those scans.
Their system, JSidentify-V2, works in four stages: it instruments the code, slices out the hot objects that run most, builds a dependency graph of memory use, and compares graphs for likeness. Memory behavior, the team argue, stays steady even when code text is scrambled. Tests against eight obfuscation methods on 1,200 mini-games back the claim. The method targets a plagiarism style that hides its own source.
Catalog record
| CREATOR | Zhihao Li 2 4 , Chaozheng Wang 2 3 4 , Zongjie Li 5 6 , Xinyong Peng 2 , Qun Xia 2 , Haochuan Lu 2 , Ting Xiong 2 , Shuzheng Gao 3 , Cuiyun Gao 3 , Shuai Wang 6 , Yuetang Deng 2 , Huafeng Ma 2 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-01655
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
JSidentify-V2: Leveraging Dynamic Memory Fingerprinting for Mini-Game Plagiarism Detection Zhihao Li 2 4 , Chaozheng Wang 2 3 4 , Zongjie Li 5 6 , Xinyong Peng 2 , Qun Xia 2 , Haochuan Lu 2 , Ting Xiong 2 , Shuzheng Gao 3 , Cuiyun Gao 3 , Shuai Wang 6 , Yuetang Deng 2 , Huafeng Ma 2 Abstract The explosive growth of mini-game platforms has led to widespread code plagiarism, where malicious users access popular games’ source code and republish them with modifications. While existing static analysis tools can detect simple obfuscation techniques like variable renaming and dead code injection, they fail against sophisticated deep obfuscation methods such as encrypted code with local or cloud-based decryption keys that completely destroy code structure and render traditional Abstract Syntax Tree analysis ineffective. To address these challenges, we present JSidentify-V2, a novel dynamic analysis framework that detects mini-game plagiarism by capturing memory invariants during program execution. Our key insight is that while obfuscation can severely distort static code characteristics, runtime memory behavior patterns remain relatively stable. JSidentify-V2 employs a four-stage pipeline: (1) static pre-analysis and instrumentation to identify potential memory invariants, (2) adaptive hot object slicing to maximize execution coverage of critical code segments, (3) Memory Dependency Graph construction to represent behavioral fingerprints resilient to obfuscation, and (4) graph-based similarity analysis for plagiarism detection. We evaluate JSidentify-V2 against eight obfuscation methods on a comprehensive dataset of 1,200 mini-games. Our framework achieves over 90% similarity detection across all tested obfuscation techniques, maintaining high accuracy even against advanced decryption-based methods where existing tools achieve near 0% detection rates. In real-world deployment, JSidentify-V2 achieves 100% precision and 99.8% recall while delivering an 8 × \times speedup compared to previous methods. Our production deployment demonstrates that plagiarism complaints have decreased by over 80%, proving JSidentify-V2’s effectiveness in protecting intellectual property rights in mini-game ecosystems. I Introduction The proliferation of smartphone technology has fundamentally transformed the mobile gaming landscape, driving unprecedented growth in this sector over the past decade. According to industry analytics, global mobile game revenue is projected to reach $135 billion across app stores [ 1 ] . This immense profitability has not only fostered a burgeoning ecosystem of game development but has also led to the proliferation of various mini-game platforms. These platforms distinguish themselves from traditional mobile game applications by embedding lightweight games directly within existing mobile applications, thereby eliminating the need for separate downloads or installations. This “plug-and-play” functionality substantially enhances user engagement and has attracted a vast user base, alongside a rapidly expanding community of developers and platform providers. A prime example of such a successful ecosystem is the Tencent WeChat mini-game platform, which has garnered hundreds of thousands of developers and hosts hundreds of thousands of mini-games, serving a user base exceeding one billion and boasting 500 million monthly active users [ 2 ] . Fig. 1 : Examples of advanced deep obfuscation techniques that static analysis fails to detect. In (a) Local Key Decryption, the core code is encrypted, and the decryption key is retrieved from a local file. In (b) Cloud Key Decryption, the key is fetched from a remote server via a network request, with decryption happening dynamically. While the platform provided by WeChat has undeniably been a pivotal force in accelerating the development and adoption of mini-games, it has concurrently introduced significant challenges, particularly concerning intellectual property infringement. The inherent characteristics of mini-games, often developed with simplified architectures and readily accessible codebases (e.g., JavaScript), render them particularly vulnerable to various forms of plagiarism. Malicious users frequently exploit the success of popular mini-games by illicitly acquiring their source code and repackaging it with modifications ranging from superficial changes such as asset replacements to sophisticated obfuscation techniques designed to evade detection. These plagiarized versions are then uploaded to the platform as competing products, directly infringing upon the intellectual property rights of original creators while diverting potential revenue streams away from legitimate developers. Such plagiarism not only undermines the creative efforts and economic interests of original developers but also poses broader threats to the platform’s ecosystem integrity and user trust. To address the escalating issue of code plagiarism in mini-games, the WeChat development team previously proposed JSidentify [ 3 ] , a framework designed for the detection of plagiarized mini-game code. JSidentify primarily leverages static analysis techniques to identify common obfuscation methods such as code restructuring applied to JavaScript code. However, recent developments over the past two to three years have revealed the emergence of significantly more challenging obfuscation techniques that effectively bypass JSidentify’s detection capabilities. Figure 1 illustrates two typical examples of these advanced obfuscation strategies: (a) local key decryption, where plagiarists encrypt the core code and employ file-based key retrieval, and (b) cloud key decryption, where decryption occurs dynamically via network communication. These deep obfuscation techniques severely corrupt the structural integrity of the code, profoundly damaging Abstract Syntax Trees (ASTs) and character-based features. Consequently, static analysis-based de-obfuscation approaches fail due to the critical loss of information, rendering them ineffective against these sophisticated attacks. In light of the limitations faced by existing static analysis methods when confronted with these sophisticated deep obfuscation techniques, this work explores the necessity and feasibility of incorporating dynamic analysis. The fundamental insight driving this approach is that while obfuscation can severely disrupt the static structural characteristics of code, the program’s runtime memory behavior patterns tend to remain relatively stable. Building upon this observation, this paper proposes a novel detection framework that synergistically combines dynamic and static analysis, designed to identify plagiarism by capturing and analyzing the program’s memory state during execution. The core concept of this methodology lies in identifying memory invariants , which are runtime artifacts whose values or relational properties remain consistent across different executions of programs sharing the same core logic, even under deep obfuscation. These memory invariants are then organized into Memory Dependency Graphs (MDGs) that serve as robust behavioral fingerprints, enabling effective plagiarism detection. Building upon the aforementioned arguments for the viability of dynamic analysis, we introduce JSidentify-V2, a new generation anti-plagiarism framework for mini-games. Our proposed framework employs a novel four-stage pipeline approach, offering a systematic solution to the complex challenge of deep obfuscation. The framework integrates four core components: (1) Static Pre-analysis and Instrumentation that identifies potential memory invariants and inserts monitoring probes; (2) Adaptive Hot Object Slicing that intelligently enhances execution coverage by focusing on critical program entities and domain-specific patterns; (3) Memory Dependency Graph (MDG) Construction that organizes collected invariant data into structured behavioral fingerprints; and (4) Graph-based Similarity Analysis that compares MDGs to detect plagiarism. The key innovation lies in our adaptive hot object slicing strategy, which systematically targets frequently executed functions and semantically important identifiers to maximize coverage of obfuscated code segments. This approach effectively captures behavioral invariants that remain stable across sophisticated obfuscation transformations, providing a robust and resilient solution to plagiarism detection. We comprehensively evaluate JSidentify-V2’s effectiveness against various obfuscation methods, demonstrating superior performance over existing approaches with over 90% similarity detection across all eight tested obfuscation techniques. Remarkably, for advanced methods such as Local Key Decryption and Cloud Key Decryption, where the previous best baseline JSidentify achieves only around 5% detection rate, our approach maintains consistently high