Using Claude to reduce the attack surface of code
Web page · https://anthropic.com/research/attack-surface-reduction
First seen by Oriel on .
AI summary
Anthropic reports using Claude to identify and reduce unnecessary code reachability across 25 open-source projects, arguing that limiting exposed code can reduce risk from unknown vulnerabilities.
- Claude inventoried more than 17,000 configuration options across projects totaling over five million lines of code, including 3,842 security-relevant options, and identified 493 issues.
- Findings included 162 undocumented security-sensitive configuration combinations; 94 silently disabled a security control, while others left supposedly disabled functionality reachable or created inconsistent behavior.
- In OpenSSH, Claude found roughly 4,500 lines of ML-DSA-44 code reachable before configuration checks and identified further gaps in the 10.5 checks; maintainers shipped fixes in version 10.6.
- The reusable Claude skill uses three passes for default reachability, execution before security gates, and risky configuration interactions, confirming findings with dynamic traces and regression tests.
- The authors say the approach reduces exposure rather than removing vulnerabilities, and changes affecting non-default configurations are left for human review.
#claude #software security #attack surface reduction #openssh
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Points
Score reported by the source.
- Hacker News: 132, latest observation ; +127 over observed period since
Comments
Comment count reported by the source.
- Hacker News: 88, latest observation ; +88 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hacker News | 50175779 | https://www.anthropic.com/research/attack-surface-reduction Read discussion |