A chain is only as strong as its weakest link, and hallucinating dependencies will damage your software supply chain. DO NOT trust blindly on AI generators.

TL;DR: Avoid hallucinated or fake packages that can compromise security and stability.

Problems 😔

• Unsupervised Vibe Coding
• Security risks
• Dependency confusion
• Too Many Imports
• Typesquatting
• Injection attacks
• Supply chain corruption
• Untrusted sources
• Package Poisoning

Solutions 😃

1. Validate package names
2. Use trusted repositories
3. Lock dependencies
4. Monitor for typos
5. Audit third-party packages
6. Lock dependency versions
7. Use private repositories
8. Verify package checksums
9. Implement allow-lists
10. Audit dependencies regularly

Context 💬

When AI generated code add external libraries to your project, you are assuming they come from reliable sources.

If you’re not careful, you might accidentally pull a malicious or incorrect package.

From Helpful to Harmful: How AI Recommendations Destroyed My OS

Maxi Contieri ・ Apr 21

#ai
#programming
#chatgpt
#refactoring

This is called “package hallucination” .

Attackers often publish fake packages with names similar to popular ones (typesquatting), hoping developers will install them by mistake.

These packages can inject harmful code into your system through the package supply chain.

In a recent paper, the authors found a lot of evidence of these attacks on the wild.

Researchers tested 16 language models and generated more than half a million code snippets.

They found that nearly 440,000 dependencies pointed to libraries that simply don’t exist.

These are very harmful backdoors for hackers.

Sample Code 📖

Wrong ❌

// package.json
{
  "name": "my-app",
  "dependencies": {
    "react": "^18.2.0",
    "lodahs": "1.0.0",  // Typosquatting attack
    "internal-logger": "2.1.0" 
    // Vulnerable to dependency confusion
  }
}

Right 👉

// package.json
{
  "name": "my-app",
  "dependencies": {
    "react": "18.2.0",
    "lodash": "4.17.21",  // Correct spelling with exact version
    "@company-scope/internal-logger": "2.1.0" // Scoped package
  },
  "resolutions": {
    "lodash": "4.17.21"  
    // Force specific version for nested dependencies
  },
  "packageManager": "yarn@3.2.0" // Lock package manager version
}

Detection 🔍

You can detect this smell by reviewing all dependencies manually and using tools like automated linters or IDEs that flag suspicious or misspelled package names.

Also, dependency lock files help track exactly which versions were installed.

Tags 🏷️

• Security

Level 🔋

Why the Bijection Is Important 🗺️

Modeling a one-to-one between real-world dependencies and those in your code ensures trust and predictability.

When you allow hallucinated packages, you break this trust, potentially introducing defects, security holes, and maintenance nightmares.

AI Generation 🤖

AI generators can unintentionally create this smell by suggesting incorrect or non-existent package names as the article proved.

They may confuse similar-sounding libraries or suggest outdated/renamed packages.

AI Detection 🥃

AI can fix this smell when given clear instructions to validate package names against official registries or enforce naming conventions.

With proper training data, AI tools can flag potential typesquatting attempts automatically.

Try Them! 🛠

Remember: AI Assistants make lots of mistakes

Suggested Prompt: verify and replace invalid packages

Conclusion 🏁

Package hallucination is a dangerous code smell that exposes your application to serious threats.

By validating every dependency and using strict version controls, you protect yourself from malicious injections and ensure software integrity.

Relations 👩‍❤️‍💋‍👨

Code Smell 138 – Packages Dependency

Maxi Contieri ・ Jun 3 ’22

#javascript
#npm
#node
#security

Code Smell 90 – Implementative Callback Events

Maxi Contieri ・ Oct 7 ’21

#javascript
#webdev
#beginners
#tutorial

Code Smell 263 – Squatting

Maxi Contieri ・ Aug 13 ’24

#webdev
#beginners
#programming
#python

More Information 📕

We Have a Package for You! A Comprehensive Analysis of Package Hallucinations
by Code Generating LLMs

Disclaimer 📘

Code Smells are my opinion.

Credits 🙏

Photo by JJ Ying on Unsplash

Controlling complexity is the essence of computer programming.

Fred Brooks

Software Engineering Great Quotes

Maxi Contieri ・ Dec 28 ’20

#codenewbie
#programming
#quotes
#software

This article is part of the CodeSmell Series.

How to Find the Stinky parts of your Code

Maxi Contieri ・ May 21 ’21

#codenewbie
#tutorial
#codequality
#beginners