AI summary by TwinMind
The crypto scam that almost ruined npm forever
Theo - t3․gg
13m 46s
NPM Spam Problem Overview
- Phylum research team discovered spam packages in npm associated with T protocol (tea.xyz)
- T protocol: decentralized initiative promising to compensate software devs in cryptocurrency for open-source contributions
- Q2 2024 estimate: approximately 1 in 4 packages published to npm were T-associated spam
- Virtually all spam packages had no redeeming quality aside from gaming protocol to inflate dev contributions artificially
T Protocol Criticism and Open Source Philosophy
- Creator's perspective: crypto solutions cannot solve open source funding problems
- Open source works because code given with no expectation of return
- Real solution: cultural shift to pressure businesses using open source to pay maintainers
- Problem: giant businesses built on open source without compensating creators
- T protocol creator is Homebrew creator (referenced Google hiring story)
Spam Package Volume and Statistics
- Baseline early 2024: ~1,500 packages ingested daily from npm (business days), ~750 on weekends
- February 2024: steady increase from few thousand to tens of thousands
- Peak: April 8, 2024 - over 48,000 packages published in single day
- Q2 2024 random sample analysis (1,600 packages manually triaged):
- 95% confidence interval: 21-25% of all packages were spam
- Estimated over 500,000 spam packages in Q2
New Package Analysis (Never-Before-Seen)
- February to present: 890,000+ new packages identified
- Sample of 900 new packages analyzed
- Results: 68-75% were spam (95% confidence interval)
- Estimated 630,000-670,000 spam packages from February to analysis date
- 5 of every 7 packages published over past 6 months are T-spam
Spam Package Characteristics
- Gibberish package names
- Random word combinations from lists
- Implausible dependent package lists
- Dubious number of dependencies
- Ubiquitous t.yaml file identifying code owner
- Recommendation: don't trust packages with t.yaml files
Security and Ecosystem Threats
- No immediate malicious code discovered in spam packages
- Potential dangers:
- Typosquatting opportunities (e.g., "web3-cover" vs legitimate names)
- AI model training contamination (garbage in, garbage out)
- Raised noise floor hindering security analysis
- Environment for hiding actual maliciousness
- Supply chain complexity through transitive dependencies
NPM Display Bug Discovery
- Example: "sour-noise" package
- NPM website incorrectly shows dependency on legitimate "axios"
- Actually depends on random tarball URL to spam package
- NPM's dependency display system fundamentally broken
- Axios incorrectly shows "sour-noise" as dependent
- Transitive dependencies can pull unwanted/unexpected packages
Remediation Efforts
- T protocol taking steps to address spam problem
- NPM removing some spammers but takedown rate < publication rate
- Problem extends beyond NPM:
- Ruby Gems: user published ~1,800 spam packages (Feb-Mar 2024)
- Phylum continuing research and detection method development
Research Source
- Investigation by Phylum (software supply chain security company)
- Comparison to CrowdStrike but addressing software origin problems
- Quarterly research reports tracking spam trends
