Auto-Detection of Copied Code from StackOverflow

on YouTube ↗25m

" summary

Claims an automated pipeline that detects code copied from StackOverflow by scraping snippets into a local NoSQL DB and running method-wise comparisons with a custom detector. It uses StackOversight to scrape, CorpusCC for tokenizing and set-similarity detection (Java tokenizer initially, ANTLR for Python), and plans a Redis queue for scaling. An initial run reports ~6,000 raw matches against Spring Framework, which filter down to about a dozen relevant copies (uploader also reports 500+ high-severity clones).

AI-written summary · about

" chapters (13)

1. 0:00Intro
2. 0:48Outline
3. 1:12What is Code Clone Detection Process of finding two pieces of code within an application that are duplicates of one another
4. 1:41Why Should You Care?
5. 5:02Types of Code Clones - Type 4
6. 5:53Types of Code Clone Detection
7. 8:20Overview of Stack Oversight
8. 13:43Overview of Corpus-CC
9. 15:00Overview of Persistence Database
10. 15:43How does it detect clones - Every method of the target application is tokenized
11. 16:24Results
12. 17:37Future Work
13. 18:23Acknowledgements

" related talks

• 29mCopy-Paste Detector for 150+ Languages on JavaScriptAndrey Kucherenko · OdessaJS 2018
• 37mMachine Learning on Source CodeFrancesc Campoy Flores · Scale By The Bay 2018
• 26mDetecting and Mapping Code Defects Introduced by Generative AI Using CodeLogic & Neo4jBrandon Tylke, Edwin Gnichtel · NODES 2024
• 49mTreat Your Code as a Crime SceneAdam Tornhill · GOTO Chicago 2016
• 25mMachine Learning for source code analysisAlexander Bezzubov · FOSSASIA Summit 2017
• 55mExploring StackOverflow dataEvelina Gabasova · NDC London 2017
• 42mContemporary Automatic Program AnalysisJulian Cohen · Black Hat USA 2014
• 23mStopping vulnerabilities at the sourcePavel Avgustinov, Grey Baker · GitHub Satellite Virtual 2020