Federal agency cyber incidents, FY 2016 to FY 2023: flat totals, shifting vectors, two missing years

We read every OMB FISMA annual report to Congress we could find to chart federal incidents by attack vector and severity, and to mark where the counting rules changed.

The question, and how far the record reaches

Each year the Office of Management and Budget (OMB) reports to Congress under the Federal Information Security Modernization Act of 2014 (FISMA), and since the FY 2016 report it has counted the incidents agencies reported to US-CERT, now CISA, by attack vector [2]. From the FY 2020 report onward it has also counted them by priority level under CISA's National Cyber Incident Scoring System (NCISS) [6]. This piece reads those tables across every year we could find, to see what changed in how federal agencies are hit and how badly.

We found nine reports, for fiscal years 2015 to 2023, all on official White House sites: the FY 2015 to FY 2018 reports on the archived 2017–2021 White House site, and the FY 2019 to FY 2023 reports on the archived 2021–2025 site [1] [2] [3] [4] [5] [6] [7] [8] [9]. We did not find a report for FY 2024 or FY 2025. We looked on whitehouse.gov (the OMB pages, the site search and the upload folders where the FY 2023 report's naming pattern would put a later one) and in web searches for each title, on November 16, 2026; the Internet Archive's index was offline when we tried it, so we could not check for a copy there. Our federal zero trust and logging outlook could not find them either. If a report for either year exists elsewhere, we have not seen it, and nothing here describes either year.

That leaves eight years of attack vector data (FY 2016 to FY 2023), five years of severity data (FY 2019 to FY 2023), and eight years of major incident counts under definitions that changed after the first of them. The question is narrowed to those years.

The OMB FISMA annual reports to Congress we found and the tables we used, as checked on November 16, 2026
ReportAttack vector tableSeverity tableMajor incidents stated
FY 2015 [1]Table 1 (totals for FY 2013 to FY 2015, legacy US-CERT categories) [1]None [1]States that no agency had reported a major incident since M-16-03 was issued [1]
FY 2016 [2]Table 6 [2]None [2]16, under M-16-03 [2]
FY 2017 [3]Table 18 (FY 2016 and FY 2017) [3]None [3]5, under M-18-02 [3]
FY 2018 [4]Table 16 (FY 2017 and FY 2018) [4]None [4]0, the first year with none, under M-19-02 [4]
FY 2019 [5]Table 16 (FY 2018 and FY 2019) [5]NCISS rating given for each major incident only [5]3, under M-20-04 [5]
FY 2020 [6]Table 6 (FY 2019 and FY 2020) [6]Table 7 (FY 2019 and FY 2020) [6]6, under M-20-04 [6]
FY 2021 [7]Table 4 (FY 2020 and FY 2021) [7]Table 5 (FY 2020 and FY 2021) [7]7, under M-21-02 [7]
FY 2022 [8]Table 3 (FY 2021 and FY 2022) [8]Table 4 (FY 2021 and FY 2022) [8]4 in the body (the key takeaways box says three), under M-22-05 [8]
FY 2023 [9]Table 3 (corrected FY 2022 and FY 2023) [9]Table 4 (corrected FY 2022 and FY 2023) [9]11, under M-23-03 [9]
FY 2024Not found on whitehouse.gov or the White House archive sites
FY 2025Not found on whitehouse.gov or the White House archive sites

Where the counting rules changed

The series cannot be read as one line from FY 2013 to FY 2023, and parts of it cannot be read as one line even after FY 2016. These are the breaks we found in the reports and in CISA's guidelines.

  • Before FY 2016, different categories. The FY 2015 report counted 60,753, 69,851 and 77,183 incidents for FY 2013, FY 2014 and FY 2015 under US-CERT's legacy categories, such as Non-Cyber, Policy Violation and Other, which included scans, probes and attempted access [1]. US-CERT's 2015 guidelines took effect on October 1, 2014 but let agencies keep the legacy categories until September 30, 2015 [11]. The FY 2016 report says its attack vector data and prior years' data are not comparable [2]. We do not chain the two.
  • Validation from late FY 2016. US-CERT and agencies began validating incident data in late FY 2016 to remove non-cyber reports, scans, probes and duplicates [2], and the FY 2017 report describes the process as begun during FY 2017 [3]. FY 2016 counts are post-validation, but the process was new.
  • The category list moved in FY 2017 and FY 2018. The FY 2017 table dropped Impersonation/Spoofing, folding FY 2016's 64 such incidents into Other (shown as 11,866), and added a Physical Cause row with 7 incidents [2] [3]. The FY 2018 table brought Impersonation/Spoofing back, renamed Other to Other/Unknown, and folded Physical Cause into it, restating FY 2017's figure as 10,825 [4]. From FY 2018 to FY 2023 the nine rows stay the same [4] [9].
  • Reporters inside and outside FISMA. The FY 2021 vector table excludes 34 incidents from non-FISMA entities (32,509), while its severity table includes them (32,543) [7]. From FY 2022 the severity tables include entities outside the federal executive branch, so they total more than the vector tables: 30,881 against 30,659 for FY 2022 [8], and 32,229 against 32,211 for FY 2023 [9].
  • A restated year. The FY 2023 report corrected FY 2022 from 30,659 to 29,319 incidents, with Other/Unknown cut from 12,489 to 11,144, and calls the change a correction to counts and categories [8] [9]. We show both and use the corrected figures when comparing FY 2022 with FY 2023.
  • Major incident definitions. FY 2016 used M-16-03, under which an incident affecting 10,000 or more records or users could qualify; M-17-05 redefined a major incident around demonstrable harm and set the breach threshold at 100,000 or more people's personal information [2]. Each later report counts major incidents under that year's memorandum, from M-18-02 to M-23-03 [3] [4] [5] [6] [7] [8] [9]. FY 2016's count is not comparable with later years.
  • Tables and text that disagree. The FY 2021 narrative gives 14,014 Other/Unknown and 9,875 Improper Usage incidents, which are the CFO Act agency columns; the government-wide figures in its table are 14,805 and 10,123, and the table's rows add to 32,511 against a stated total of 32,509 [7]. The FY 2022 key takeaways box says three major incidents; the body says four [8]. We use the government-wide table column and the body text.
  • The next change is announced. CISA says its 2017 Federal Incident Notification Guidelines, whose attack vector taxonomy the OMB reports say they follow [4], are being updated to support a new cyber incident reporting form published on November 9, 2026, and will be renamed the Federal Incident Reporting Requirements [10].

The attack vectors themselves are CISA's: Unknown, Attrition, Web, Email/Phishing, External/Removable Media, Impersonation/Spoofing, Improper Usage, Loss or Theft of Equipment, and Other [10]. The OMB tables merge Unknown with Other and add a row for incidents using two or more vectors [4]. Improper Usage means a violation of acceptable use policies by an authorized user, such as installing file-sharing software that leaks sensitive data [10]; it is not an outside attack.

Incidents by attack vector, FY 2016 to FY 2023

Government-wide incidents by attack vector, as first published in each year's OMB FISMA report (FY 2021 excludes 34 incidents from non-FISMA entities) [2] [3] [4] [5] [6] [7] [8] [9]. A dash means the category was not in that year's table.
Attack vectorFY 2016FY 2017FY 2018FY 2019FY 2020FY 2021FY 2022FY 2023
Attrition109 [2]151 [3]163 [4]332 [5]345 [6]440 [7]197 [8]1,147 [9]
Email/Phishing3,292 [2]7,328 [3]6,930 [4]4,388 [5]4,264 [6]2,962 [7]3,010 [8]6,198 [9]
External/Removable Media138 [2]72 [3]32 [4]47 [5]32 [6]15 [7]47 [8]92 [9]
Impersonation/Spoofing64 [2]–47 [4]35 [5]93 [6]272 [7]35 [8]30 [9]
Improper Usage4,130 [2]7,856 [3]9,674 [4]12,507 [5]11,874 [6]10,123 [7]10,467 [8]12,261 [9]
Loss or Theft of Equipment5,690 [2]4,395 [3]2,552 [4]1,885 [5]1,242 [6]1,071 [7]1,786 [8]3,135 [9]
Web4,868 [2]4,049 [3]3,332 [4]1,982 [5]2,753 [6]2,730 [7]2,453 [8]3,569 [9]
Physical Cause–7 [3]––––––
Other (Other/Unknown from FY 2018)11,802 [2]10,818 [3]8,285 [4]7,240 [5]10,102 [6]14,805 [7]12,489 [8]5,687 [9]
Multiple Attack Vectors806 [2]601 [3]92 [4]165 [5]114 [6]93 [7]175 [8]92 [9]
Total30,899 [2]35,277 [3]31,107 [4]28,581 [5]30,819 [6]32,509 [7]30,659 [8]32,211 [9]
Share of each year's published total, our arithmetic from the tables above (percent, one decimal)
Attack vectorFY 2016FY 2017FY 2018FY 2019FY 2020FY 2021FY 2022FY 2023
Improper Usage13.4% [2]22.3% [3]31.1% [4]43.8% [5]38.5% [6]31.1% [7]34.1% [8]38.1% [9]
Email/Phishing10.7% [2]20.8% [3]22.3% [4]15.4% [5]13.8% [6]9.1% [7]9.8% [8]19.2% [9]
Loss or Theft of Equipment18.4% [2]12.5% [3]8.2% [4]6.6% [5]4.0% [6]3.3% [7]5.8% [8]9.7% [9]
Web15.8% [2]11.5% [3]10.7% [4]6.9% [5]8.9% [6]8.4% [7]8.0% [8]11.1% [9]
Other or Other/Unknown38.2% [2]30.7% [3]26.6% [4]25.3% [5]32.8% [6]45.5% [7]40.7% [8]17.7% [9]

Totals sat in a narrow band

Across the eight years, the published total ranged from 28,581 in FY 2019 to 35,277 in FY 2017 [3] [5]. The largest year-on-year move was the 14% rise from FY 2016 to FY 2017 [3] [4] [5] [6] [7] [8] [9]; the FY 2023 rise was 9.9% on the corrected FY 2022 figure [9]. OMB itself cautioned that drawing conclusions from a single year's change, while agencies adjusted to new reporting guidelines, would be concerning (the FY 2018 report) or premature (the FY 2022 report) [4] [8].

Moraph's view

Moraph's view: a band of roughly 28,500 to 35,300 a year, under rules that kept changing, does not show federal agencies being hit more or less often. The totals measure what agencies detect, validate and report; in the FY 2023 report, agencies credit better detection, automation, training and changed tracking methods for that year's increase. The composition is the more useful signal.

Improper Usage became the largest named vector and stayed there

In FY 2016 the largest named vector was Loss or Theft of Equipment, with 5,690 incidents [2]. From FY 2017 to FY 2023 Improper Usage was the largest vector other than Other/Unknown in every year, peaking at 12,507 in FY 2019 and standing at 12,261, or roughly 38%, in FY 2023 [3] [5] [9]. OMB reads its prevalence as a sign that agencies detect policy violations but lack automated enforcement or prevention [5] [9].

Unknown vectors rose, then halved

Incidents filed as Other, or from FY 2018 Other/Unknown, made up 38.2% of the FY 2016 total and 45.5% of FY 2021's, on our arithmetic from the tables [2] [7]. In FY 2023 they fell to 5,687, or 18% of the total, from 11,144 (38%) in the corrected FY 2022 figures [9].

Moraph's view

Moraph's view: the drop in unknown vectors is the largest single change in the series, and we think much of FY 2023's rise in named vectors is likely the same kinds of incident being classified, not new ones; OMB reads the fall as agencies getting better at detecting and categorizing attacks. When a category halves in one year, the categories around it cannot be read as independent trends.

Phishing and lost equipment came back in FY 2023

Email/Phishing incidents peaked at 7,328 in FY 2017, fell to 2,962 by FY 2021, and doubled from 3,011 in the corrected FY 2022 figures to 6,198 in FY 2023 [3] [7] [9]. Loss or Theft of Equipment fell from 5,690 in FY 2016 to 1,071 in FY 2021 and rose to 3,135 in FY 2023 [2] [7] [9]. Attrition, which covers brute force attacks and denial of service, went from 197 to 1,147 between the corrected FY 2022 and FY 2023 figures [9] [10]. Web incidents moved between 1,982 and 4,868 a year without a steady direction [2] [5].

Severity: nearly everything is baseline

CISA assigns every reported incident an NCISS priority level, from Baseline – Negligible (an unsubstantiated or inconsequential event) up to Emergency [10]. OMB began publishing the spread in the FY 2020 report, with FY 2019 as the comparison year, and says the scores are a relative mechanism for prioritization, not an absolute measure of risk [6].

Incidents by NCISS priority level, from the OMB report tables that carry them (Table 7 of the FY 2020 report, Table 5 of FY 2021, Table 4 of FY 2022 and FY 2023) [6] [7] [8] [9]. Severe (Red) and Emergency (Black) were zero in every year shown [6] [7] [8] [9]. Totals here include non-FISMA and non-executive-branch reporters where the report says so.
Priority levelFY 2019FY 2020FY 2021FY 2022 as first publishedFY 2022 as correctedFY 2023
Uncategorized471 [6]177 [6]2,384 [7]1,670 [8]263 [9]229 [9]
Baseline – Negligible (White)19,850 [6]19,039 [6]16,783 [7]16,511 [8]16,449 [9]14,086 [9]
Baseline – Minor (Blue)7,316 [6]10,765 [6]12,766 [7]12,205 [8]12,139 [9]17,535 [9]
Low (Green)938 [6]831 [6]593 [7]493 [8]491 [9]348 [9]
Medium (Yellow)6 [6]7 [6]14 [7]2 [8]1 [9]31 [9]
High (Orange)0 [6]0 [6]3 [7]0 [8]0 [9]0 [9]
Total28,581 [6]30,819 [6]32,543 [7]30,881 [8]29,343 [9]32,229 [9]
Low (Green) or above, our arithmetic944 [6]838 [6]610 [7]495 [8]492 [9]379 [9]

The two baseline levels made up about 95% of FY 2019 incidents and about 98% of FY 2023's, on our arithmetic from the tables; OMB's text puts the baseline share at approximately 97% for both FY 2019 and FY 2020, and approximately 99% for FY 2023 [6] [9]. Incidents rated Low (Green) or above fell each year, from 944 in FY 2019 to 379 in FY 2023 [6] [9]. High (Orange) appears only in FY 2021, with 3 incidents, and no year shows Severe or Emergency [6] [7] [8] [9]. Medium (Yellow) went from 1 in the corrected FY 2022 figures to 31 in FY 2023 [9]. FY 2021's 2,384 uncategorized incidents, which OMB attributes partly to requests for information and abuse notifications that received no score, make that year the hardest to compare [7].

Major incidents

Agencies designated 16 major incidents in FY 2016, 5 in FY 2017, none in FY 2018, 3 in FY 2019, 6 in FY 2020, 7 in FY 2021, 4 in FY 2022 and 11 in FY 2023 [2] [3] [4] [5] [6] [7] [8] [9]. The FY 2016 figure used the broader M-16-03 definition [2]. Several of FY 2021's were tied to the SolarWinds compromise, which the FY 2020 report left out because it was first reported in December 2020 [6] [7]. In FY 2023 the Departments of Health and Human Services, Justice and the Treasury each reported more than one; one HHS incident was a ransomware attack on a contractor-operated system that exposed data on over 2.8 million people [9].

Moraph's view

Moraph's view: the severity data and the major incident counts point in different directions in FY 2023 (fewer incidents rated Low or above, more designated major), and both can be true. A major incident is the agency's own call against an OMB definition that leans on harm and on breaches of personal information, while NCISS scores national-level impact. Several FY 2023 major incidents were breaches at contractors, which an agency's own network telemetry would not show. We would not read either line as a trend until a report for FY 2024 or later is published.

What we left out, and why

  • FY 2024 and FY 2025: no OMB report found, as set out above. We did not fill the gap from CISA dashboards, inspector general audits, GAO reports or press coverage, because the question is what OMB's reports show.
  • Per-agency counts: each report has agency summaries [2] [9], but agency-level vectors are outside this piece.
  • FY 2013 to FY 2015 by category: the FY 2015 report's figures use the legacy categories [1] and are shown only as totals, for context.
  • Breach counts reported through the privacy metrics, which some reports also give [7].

Moraph's view

Moraph's view: the most important finding may be the gap. FISMA asks OMB for an annual report, and from FY 2015 to FY 2023 one was published every year. As of November 16, 2026, the public record of federal incidents by vector and severity stops at September 30, 2023, while the rules for counting them are about to change again with CISA's new reporting requirements. When the series resumes, the first new year may need its own break in the line.

What this means for you

  • Federal agency CISOs and security operations leads: Expect the incident fields you send CISA to change: CISA says its guidelines are being rewritten as the Federal Incident Reporting Requirements to support a reporting form published on November 9, 2026 [10]. OMB read FY 2023's halving of unknown vectors as agencies getting better at detecting and categorizing attacks [9]; the vector field is the one the public tables are built from.
  • Contractors and service providers to agencies: Several FY 2023 major incidents began at contractors, including a ransomware attack on a contractor-operated system supporting CMS and a zero-day exploited at two HHS contractors [9], and a FY 2019 major incident followed ransomware at a CBP contractor [5]. FISMA incident reporting covers agency systems whether managed by the agency, a contractor or another source, so a contractor breach can become the agency's major incident [10].
  • Oversight staff, auditors and researchers using these counts: Do not join FY 2015 and earlier totals to the attack vector series [2]; use the corrected FY 2022 figures when comparing with FY 2023 [9]; and note that the severity tables include reporters outside the executive branch from FY 2022 [8] [9]. Read the table, not the summary box, where the two disagree [7] [8].

Moraph predictions

Every prediction is Moraph's view, not a fact. Each has a date, a confidence level (70% means we expect to be wrong about three times in ten) and a test anyone can apply. We score them on this page when the date comes and never remove one. How we make and score predictions.

  1. P1 By Confidence 60% Open

    No OMB annual FISMA report to Congress covering fiscal year 2024 or any later fiscal year is posted on whitehouse.gov or an official White House archive site by December 31, 2027.

    How we will check: On or after January 1, 2028, search whitehouse.gov and the White House archive sites for a document titled as a FISMA Annual Report to Congress whose cover names fiscal year 2024 or later. Right if none was posted on or before December 31, 2027 (by its page date or upload folder); wrong if one was.

    Basis: OMB published a report for each of fiscal years 2015 to 2023 [1] [9], each issued under 44 U.S.C. § 3553 [2]. As of November 16, 2026 we found none for FY 2024 or FY 2025, so more than two years have passed since FY 2024 ended without one.

  2. P2 By Confidence 55% Open

    OMB posts at least one annual FISMA report to Congress covering fiscal year 2024 or a later fiscal year by December 31, 2028.

    How we will check: On or after January 1, 2029, search whitehouse.gov and the White House archive sites for a FISMA Annual Report to Congress whose cover names fiscal year 2024 or later. Right if at least one was posted on or before December 31, 2028.

    Basis: The reports say they are published in accordance with FISMA, 44 U.S.C. § 3553 [2], and OMB issued one for every fiscal year from 2015 to 2023 [1] [9]; against that, none has appeared for FY 2024 or FY 2025 as of November 16, 2026.

  3. P3 By Confidence 50% Open

    By December 31, 2028, OMB posts an annual FISMA report to Congress for fiscal year 2024 or later whose attack vector table shows Improper Usage as the largest vector, other than Other/Unknown and Multiple Attack Vectors, for the most recent fiscal year it covers.

    How we will check: Find the most recent OMB FISMA Annual Report to Congress posted by December 31, 2028 that covers fiscal year 2024 or later, and read the government-wide column of its incidents-by-attack-vector table for the latest year. Right if Improper Usage is the largest named vector; wrong if another named vector is larger, if the report has no attack vector table, or if no such report is posted by then.

    Basis: Improper Usage was the largest named vector in every year from FY 2017 to FY 2023, with 12,261 incidents in FY 2023 against 6,198 for Email/Phishing, the next largest [3] [9]. The confidence also carries the chance that no report appears (P2), and that CISA's new Federal Incident Reporting Requirements change the categories [10], so it sits below P2's 55.

Facts on this page link to their sources. Text marked "Moraph's view", and every prediction, is our own analysis, not a fact or advice. Methodology.

Spotted an error? Tell us and we will check it and correct the piece where needed.

  • FISMA
  • OMB
  • CISA
  • federal agencies
  • attack vectors
  • NCISS
  • major incidents