State of EPSS and What to Expect from Version 4
summary
EPSS v4 claims to predict the probability a CVE sees exploitation in the next 30 days and expands sources to include malware, GitHub, Shodan and commercial feeds; the upgraded pipeline uses ~2,500 features and retrains more often. Training covers 82 billion CVE‑day rows, finds only 6% exploited of published CVEs, and shows strong lift at the top percentiles. Model outputs a 30-day probability plus a percentile rank; v4 improves high‑end accuracy but needs monitoring for midrange drift.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 10:16EPSS model accuracy
- 20:08EPSS version history
- 30:42Training the EPSS model
- 36:48EPSS probability explained
- 43:11EPSS key features
- 47:03Roadmap and Q&A
related talks
40mHow EPSS Is Wrong and Useful at the Same Time
01h 23mEPSS: Challenges and Opportunities Going Forward + EPSS AMA
54mTowards Real World Cyber Risk (Panel)
58mWhy Can't We All Just Get Along? Bridging the Gap in Vulnerability Prioritization Standards
15mKeynote
01h 00mFIRSTCON21 - June 8 | Stream 2 Part 2 (14:30-15:30 UTC)
41mIntegrating EPSS and CVSS in Open Policy Agent To Quarantine Real-world Vulnerabilities
32mUnderstanding Exploitability with VEX, EPSS, and Other Standard Frameworks