Investigators measured how quickly publicly disclosed software vulnerabilities were confirmed exploited. They linked entries from the CISA Known Exploited Vulnerabilities catalog to CVE publication dates and computed time from disclosure to confirmed exploitation. The cohort of catalog entries used for this linkage numbered 1,748; the source provided that count and no additional sampling details were reported. Data indicated a notable contraction in the interval after 2021. For vulnerabilities published in late 2021 through mid 2022, the median time to confirmed exploitation was 24 days. That median fell to 0 days for CVEs published in the second half of 2022, and then was 6 days across 2023–2025. The authors reported no further shortening of that interval within 2023–2025 (P = .40). A striking detail: roughly one-quarter of exploited vulnerabilities were used on or before the day they were publicly disclosed, which implies many exploits were effectively immediate.
The study team did not report detailed breakdowns by vulnerability type or vendor in the abstract. The source material specified linkage of publicly listed CVEs to the CISA catalog but did not provide per-item methodological specifics within the abstract text available here.
The authors also examined 8,029 breach reports submitted to the HHS Office for Civil Rights Breach Portal and restricted analysis to hacking/IT incidents at health care providers. They tracked how the size distribution of breaches—measured by number of affected individuals—shifted from 2016 through 2025. In 2016, 42% of provider hacking breaches affected fewer than 5,000 people. That share declined to 32% in 2022 (trend P = .004), indicating a period during which attacks concentrated more often on larger organizations. After 2022 the pattern reversed: the share of smaller breaches rose to 46% in 2025 (P < .001).
The study abstract reports these proportions and trend P values but does not include additional stratified tables or case-level illustrative examples in the abstract. The authors interpreted the post-2022 uptick in smaller events as a shift away from concentrating on very large targets.
This was a retrospective analysis using only publicly available, aggregate data sources. The investigators linked the CISA catalog to CVE publication dates from the CVE List V5 to compute time-to-exploitation measures by half-year of CVE publication (2021–2025). For breach-size trends they analyzed incident reports from the HHS Office for Civil Rights portal; the abstract notes a total of 8,029 breach reports, with a focus on hacking/IT incidents at health care providers. The abstract describes the broad analytic approach (linkage and temporal comparison) but does not detail model specifications, inclusion/exclusion criteria beyond the data sources, or adjustments for reporting delays within the abstract text.
Processed data and analysis code are available from the corresponding author on request, according to the source. The raw public sources named were the CISA catalog, the CVE Program V5 listings, and the HHS OCR Breach Portal.
Data indicated the disclosure-to-exploitation window had narrowed to on the order of days by 2023–2025, following an abrupt contraction around late 2021–2022. The authors concluded that, even before widespread generative AI availability, that window had shrunk to about one week on average. They warned that defenses relying primarily on patching speed will not adequately protect smaller providers, because a sizable fraction of exploits occur immediately or within days of disclosure. The abstract states this as a practical consequence of the timelines and breach-size changes observed.
The study did not report operational guidance or specific defensive protocols in the abstract. Nor did it provide cost or resource estimates for alternative protective measures. Those operational details were outside the abstract text.
The authors declared all data used were publicly available and aggregate; no individual-level protected health information were included. They noted the study used only data from the CISA Known Exploited Vulnerabilities catalog, the CVE List V5, and the HHS Office for Civil Rights Breach Portal. The team reported that processed data and analysis code can be obtained from the corresponding author on request. All authors are employees of Centralmedience Co., Ltd., and the authors declared no external funding. The preprint status was explicitly stated: this report has not been peer reviewed and should not be used to guide clinical practice.
The abstract documents several study limitations implicitly: reliance on public catalogs and breach reports, which may be affected by reporting delays, variable completeness, and the inclusion criteria used by the original public databases. The abstract did not provide further methodological detail or stratified sensitivity analyses in the text presented here. Requests for the full analytic code and processed datasets are directed to the corresponding author, per the data availability statement.
Unanswered questions that remain in the source: how exploit timelines vary by vulnerability class or vendor, what operational countermeasures most effectively shield small providers, and how reporting lag in public breach registries affects trend estimates. These items are not resolved in the abstract and would require access to the full manuscript or underlying code and data to evaluate further.