Know the exploit before it knows you.
Search real CVEs and public exploits, each scored by a local AI model for exploitability, blast radius, and remediation. Free, no signup, no external providers.
21,350
CVEs indexed
5,165
AI-analysed
1,641
Critical
7,798
With public exploit
Search the live archive
Full search & filters →12 of 12 CVEs match.
Showing the 12 newest of 21,350 records — type to search the full archive.
- 8.8CVE-2026-65220HIGHAI analysis pending — metadata only.
- 7.4CVE-2026-65219HIGHAI analysis pending — metadata only.
- 4.9CVE-2026-65210MEDIUMAI analysis pending — metadata only.
- 3.1CVE-2026-18451LOWAI analysis pending — metadata only.
- 9.8CVE-2026-18450CRITICALThe BMC fails to enforce integrity and encryption on IPMI commands, allowing unauthenticated attackers to send forged packets that execute arbitrary commands on an active session.
- 7.5CVE-2026-18449HIGHAI analysis pending — metadata only.
- 7.2CVE-2026-16521HIGHAI analysis pending — metadata only.
- —CVE-2026-111461UNSCOREDAI analysis pending — metadata only.
- —CVE-2026-111460UNSCOREDAI analysis pending — metadata only.
- —CVE-2026-111459UNSCOREDAI analysis pending — metadata only.
- 6.4CVE-2026-10511MEDIUMAI analysis pending — metadata only.
- —CVE-2026-107472UNSCOREDAI analysis pending — metadata only.
Data: NVD. Analysis: a local $0 model — 5,165 of 21,350 entries carry a model-written read; the rest ship metadata only and are labelled as such.
What the AI read looks like
- Summary
- A stack-based buffer overflow in the Device-Information CGI (infostat.cgi) of TOTOLINK A950RG firmware v4.1.2cu.5268_B20221104 allows remote attackers to overwrite memory via the CONTENT_LENGTH argument, enabling full system compromise due to lack of input validation and stack protections.
- Exploitability
- Trivial to exploit remotely without authentication or user interaction, as the vulnerable endpoint is publicly accessible and the attack vector (manipulating CONTENT_LENGTH) is straightforward.
- Blast radius
- Complete device takeover is likely, enabling full disclosure of sensitive data, unauthorized configuration changes, and potential use as a pivot point to compromise internal networks.
- Detection
- Look for HTTP POST requests to /infostat.cgi with abnormally large or malformed CONTENT_LENGTH headers exceeding typical device-information request sizes (e.g., >1024 bytes), especially when accompanied by unusual response behavior such as connection resets or service restarts.
- Remediation
- Upgrade to firmware version 4.1.2cu.5268_B20230101 or later if available; otherwise, immediately disable the infostat.cgi endpoint by removing execute permissions or blocking access via firewall rules until a patch is applied.
Generated locally by garnet-coder. No external provider saw this query.
Unified CVE + exploit search
One query across the CVE corpus and public proof-of-concept exploits — no more tab-hopping between six databases.
AI risk analysis
Analysed entries carry a model-generated read: exploitability, prerequisites, likely blast radius, and a prioritized remediation note.
Free API & feed
Every entry is also a static JSON endpoint, plus an RSS feed — no key, no rate limit. Read the docs.
Pricing
Search, the archive, the API and the feed are free today. Prices below are founding-member preview pricing for the paid tier and may change at launch.
Free
$0
- Full CVE & exploit search
- AI risk analysis on analysed entries
- JSON API & RSS feed
Pro
$29/mo
- Watchlists & change alerts
- Analysis on every entry, on ingest
- Bulk export
- Priority intel updates
Team
Contact us
- Everything in Pro
- Shared watchlists & seats
- SSO & audit log
- Dedicated support
Want Pro when it lands?
Search and the API stay free. Join the waitlist for founding-member pricing on watchlists, alerts, and full-archive analysis.