Medical Device Vulnerabilities: 536 CVEs, 12 on CISA's Exploited List

CISA's medical advisories directly name 536 distinct medical device vulnerabilities in 192 notices since 2016. Twelve (2.2%) are on CISA's exploited list, all Microsoft or Apache flaws, according to The PenTest Index's October 2026 analysis of CISA advisories. The often-quoted "6.2 per device" figure appears in a 2018 vendor blog post.

CISA is the U.S. cyber defense agency. A vulnerability is a security flaw. A CVE is a public ID for a disclosed flaw. Our exploited count means the flaw is on CISA's Known Exploited Vulnerabilities list as of October 8, 2026. Being off that list does not prove a flaw has never been used.

This page counts CVE IDs printed on CISA's advisory pages. One linked maker table adds 162 more IDs. It is not a safety rating for any device.

Key medical device vulnerability statistics

• 536 vulnerabilities, 192 advisories. CISA published 192 medical device security advisories from March 29, 2016 to October 8, 2026. They directly name 536 different vulnerabilities (CVEs). (The PenTest Index analysis of CISA ICS medical advisories.)
• 12 of 536 (2.2%) are on CISA's exploited list. Twelve are on CISA's Known Exploited Vulnerabilities list as of October 8, 2026. Eleven are Microsoft flaws and one is an Apache Tomcat flaw. None of these 12 catalog matches is a flaw in a device maker's own code. (The PenTest Index analysis of CISA medical advisories through October 8, 2026.)
• 0 of 77. No CISA medical advisory with a 2022 to 2026 number lists a CVE on CISA's exploited list as of October 8, 2026. The 12 sit in six advisories from 2017 to 2021. (The PenTest Index analysis of CISA medical advisories through October 8, 2026.)
• 49.4% are high or critical. Of the 512 vulnerabilities with a single CVSS v3 score, 253 are rated high or critical, and 93 (18.2%) are critical. (The PenTest Index analysis of CISA medical advisories through October 8, 2026.)
• 39.2% are access control flaws. 210 of the 536 fall in MITRE's access control group. 152 (28.4%) are authentication flaws: weak checks of a claimed identity. 42 (7.8%) are hard-coded or default passwords and keys. (The PenTest Index analysis of CISA medical advisories through October 8, 2026.)
• 76.8% need no prior privileges. 393 of the 512 scored vulnerabilities do not require the attacker to hold access rights before the attack. 257 (50.2%) can be reached over a network. (The PenTest Index analysis of CISA medical advisories through October 8, 2026.)
• Four makers, 45.8% of advisories. Philips (41), BD (23), Medtronic (15) and Baxter (9) are named in 88 of 192 advisories. In 2025 and 2026 they are named in 2 of 39. (The PenTest Index analysis of CISA medical advisories through October 8, 2026.)
• Imaging software: 10.0% to 48.4%. Imaging software was the subject of 13 of 130 advisories in 2016 to 2022 and 30 of 62 in 2023 to 2026. (The PenTest Index analysis of CISA medical advisories through October 8, 2026; device groups are our own sorting.)
• Two older reports list 1,418 and 460 flaws. Two 2016 CISA advisories report these counts for outside software in two products without listing each CVE. Their sum is 1,878, but overlap is unknown, so it is not a count of 1,878 different flaws. (CISA advisories ICSMA-16-089-01 and ICSMA-16-196-01; sum by The PenTest Index.)
• 11 of FDA's 18. Eleven of the 18 notices in the FDA's cybersecurity safety communications table (June 2013 to January 2025) match a CISA medical advisory in this dataset. (The PenTest Index matching, checked October 9, 2026 (UTC).)
• 74.5% credit an outside or anonymous reporter. 143 of 192 advisories credit an outside or anonymous reporter with finding or reporting the flaw. 49 credit only the maker. (The PenTest Index reading of CISA medical advisories through October 8, 2026.)
• The original "6.2 vulnerabilities per medical device" post gives no method. The earliest source we found is a February 7, 2018 Sensato blog post that gives no sample, dates or method. The FBI repeated the figure in 2022 and the GAO repeated the FBI in 2023.

Source for our counts: The PenTest Index analysis of CISA ICS medical advisories and the CISA Known Exploited Vulnerabilities catalog (version 2026.10.08). Data as of October 8, 2026.

The count at a glance

Source: The PenTest Index analysis of CISA ICS medical advisories and the CISA Known Exploited Vulnerabilities catalog, October 8, 2026.

How many medical device vulnerabilities are there?

CISA's medical advisory pages directly name 536 medical device vulnerabilities in 192 advisories, from March 29, 2016 to October 8, 2026. That is about 18 advisories a year. The count covers flaws CISA has published, not every flaw that exists.

Think of an advisory as a security warning for software. One notice can cover one flaw or many. The biggest one here lists 35. Of the 188 notices that name CVEs, almost half (88) list just one.

How many advisories does CISA publish each year?

The busiest year was 2018, with 32 advisories. The quietest full year was 2023, with 10. In 2025 the count jumped back to 24, and 15 of those were about imaging software.

CISA medical advisories by year, 2016 to October 8, 2026

Source: The PenTest Index analysis of CISA ICS medical advisories, October 8, 2026. Year is the year in CISA's advisory number. Fifteen CVEs appear in two advisories, so the yearly listings add to 551, not 536. Device groups are our own sorting.

What counts as one vulnerability?

One CVE counts as one vulnerability, no matter how many advisories repeat it. That is why the total is 536 and not 551.

• Advisory: one notice from CISA. CISA calls these ICS medical advisories. Each has a number like ICSMA-25-030-01: the year, the day of the year, and a counter.
• CVE: the public ID for one flaw.
• Four advisories list no CVE numbers at all. They describe groups of flaws instead. They are counted in the 192.
• The series starts in 2016. A few older medical notices carried a different kind of number and are not here.

How many medical device vulnerabilities are used in real attacks?

Twelve in this dataset are on CISA's exploited list. Of the 536 vulnerabilities, 12 (2.2%) are on CISA's Known Exploited Vulnerabilities list as of October 8, 2026. Eleven are Microsoft flaws and one is an Apache Tomcat flaw. None of these 12 catalog matches is a flaw in a device maker's own code.

CISA keeps a list of flaws that attackers have really used. A flaw gets on it only when three things are true: it has a CVE number, there is reliable evidence it has been actively exploited, and there is a clear action to address the flaw. CISA includes attempted as well as successful exploitation. Scanning, research and proof-of-concept code do not count.

Which 12 are known exploited?

All 12 are old flaws in common software: ten in Windows, one in Microsoft Office and one in Apache Tomcat. They show up in six advisories for imaging systems, a patient monitoring receiver and a drug-mixing system.

The 12 known exploited vulnerabilities named in CISA medical advisories

Source: The PenTest Index match of CISA ICS medical advisories against the CISA Known Exploited Vulnerabilities catalog, version 2026.10.08 (1,739 entries), October 8, 2026.

Three things stand out.

• The 12 sit in six advisories, all from 2017 to 2021.
• No advisory numbered 2022 or later lists one (0 of 77).
• The oldest carries a 2008 CVE number. CISA added it to its list on May 20, 2026.

One more flaw sits just outside this count. CVE-2023-43208, in NextGen Healthcare's Mirth Connect, has been on CISA's exploited list since May 20, 2024. No CISA medical advisory names it, so it is not one of the 536. CISA's September 2026 advisory for Mirth Connect (ICSMA-26-253-01) lists three other flaws.

Why are all 12 in borrowed software?

Many medical devices use common computer software. A device can inherit a flaw in the software it uses. All 12 catalog matches here are in that outside software.

• CISA's Siemens advisory says the flaws are in imaging products "running on Windows XP".
• CISA's Spacelabs advisory names the flaw BlueKeep and says an exploit "would be capable of rapidly spreading like the WannaCry malware attacks of 2017".
• One 2016 advisory says a hospital supply cabinet (the CareFusion Pyxis SupplyStation) had 1,418 known flaws in 7 outside software packages. Another says a Philips system had 460: 272 in five software packages and 188 in Windows XP.
• Add those reported counts: 1,418 + 460 = 1,878. The two lists may overlap with each other and with the 536 named CVEs, so we cannot call these 1,878 extra, different flaws. CISA's two pages put 715 and 360 at a score of 7.0 or higher: a reported sum of 1,075. These 2016 totals are kept apart from our CVSS v3 analysis.

Picture a house where the builder bought the front door lock from someone else. If that lock has a known trick, every house with that lock has the trick.

What does 12 of 536 not prove?

Twelve is the number of matches to CISA's catalog. It is not proof that the other 524 have never been used. CISA adds a flaw only with reliable evidence and a clear action to address it, so a quiet attack on one device may never make the list.

• It does not mean devices are safe. Companies that scan hospital networks find known exploited flaws on real devices. Claroty reported in March 2025 that 9% of the connected medical devices it analyzed carried a known exploited flaw, across 99% of the 351 organizations in its sample.
• Both numbers can be true. Ours counts flaws CISA named. Claroty's counts devices on hospital networks. In our data, the catalog matches are old Windows, Office and Tomcat flaws in device software.
• One advisory hides a longer list. The Biosense Webster Carto 3 advisory (ICSMA-18-107-02) prints no CVE numbers. It points to a Johnson & Johnson table instead. We checked the original Johnson & Johnson PDF. It has 183 different CVE IDs, 21 already in our data and 162 more. Add those IDs and the set is 698, with 20 (2.9%) on CISA's exploited list: 19 Microsoft and one Apache. The pattern holds. The 183-row source file gives each PDF page and the match. Rows with only Microsoft bulletin or update numbers are not turned into extra CVEs.

Has a hacked medical device hurt a patient?

The FDA says it is not aware of one in the cases it lists. Its cybersecurity page lists 18 safety notices from June 2013 to January 2025 and says: "In each of the following cases, the FDA is not aware of any patient injuries or deaths associated with cybersecurity incidents."

That sentence covers those 18 cases. It is not a promise about every device.

• The GAO, a federal watchdog, wrote in December 2023: "Although cyber incidents impacting medical devices have occurred, they are not common."
• The FDA's 2026 guidance also warns: "Cyber incidents have rendered medical devices and hospital networks inoperable, disrupting the delivery of patient care."
• The FDA's recall database lists 17 recall events that give "cybersecurity" in the reason, started between February 2019 and September 2026. This keyword search is not a full count of cybersecurity recalls or attacks. It returns 60 product records; we counted the 17 different recall event numbers.
• Two other recall records name malware: a 2008 recall of an ultrasound imaging system warned of possible computer-worm infection; a 2021 recall of cloud-hosted cancer-care software gives its reason as "Malware attack".

How serious are medical device vulnerabilities?

About half are serious on paper. Of the 512 vulnerabilities with a single score, 253 (49.4%) are rated high or critical, and 93 (18.2%) are critical, the top band.

CVSS, the Common Vulnerability Scoring System, rates a flaw from 0 to 10. Not every flaw has a published score. Think of it like a storm rating. It tells you how strong the storm could be. It does not tell you whether it will hit your house.

Medical device vulnerabilities by severity (CVSS v3)

Source: The PenTest Index analysis of scores printed in CISA ICS medical advisories, October 8, 2026. Bands follow FIRST's CVSS v3.1 scale. Twenty-four vulnerabilities in one 2018 advisory have only a score range and are left out. Shares are rounded one at a time, so 18.2% and 31.3% look like 49.5%, but 253 of 512 is 49.4%.

Here is the odd part. Of the 93 critical flaws, 88 (94.6%) are not on the known exploited list. A scary score and a real attack are two different things.

How would an attacker reach a medical device?

Half have a network attack path. Of the 512 scored vulnerabilities, 257 (50.2%) can be reached over a network, and 393 (76.8%) need no prior privileges: access rights held before the attack.

Where the attacker has to be

Source: as stated in the linked source line or in each row.

What the attack requires

Source: as stated in the linked source line or in each row.

Source for both tables: The PenTest Index analysis of CVSS v3 vectors printed in CISA ICS medical advisories, October 8, 2026. A vector records the conditions used to score a flaw. The FIRST definitions do not say every network-rated device is exposed to the public internet, or that a low-complexity attack is easy for anyone to carry out.

What do the scores leave out?

A score rates the software flaw. It does not rate what could happen to a patient. A flaw in a picture viewer and a flaw in an insulin pump can get the same number.

FIRST, the group that runs CVSS, says the base score measures severity, not risk.

What are the most common medical device vulnerabilities?

Access control flaws. Of the 536 vulnerabilities, 210 (39.2%) are access control flaws, which means the product does not check well enough who you are or what you are allowed to do. That is about 4.5 times the next kind.

The seven most common kinds of medical device vulnerability

Source: The PenTest Index analysis. Weakness labels (CWEs) are the ones CISA printed. Groups are MITRE's CWE view 1400, not ours. Thirteen smaller groups hold 102 more vulnerabilities, and 17 use CWE category IDs that we keep ungrouped. Checked October 9, 2026 (UTC).

How many are authentication and password flaws?

More than one in four. Of the 536 vulnerabilities, 152 (28.4%) are authentication flaws: the product does not check a claimed identity well enough. That includes logins, certificates and device identities. Forty-two (7.8%) are hard-coded or default passwords and keys.

The three sets nest like boxes: 210 access control flaws, 152 authentication flaws inside those, and 42 built-in passwords or keys inside those.

A built-in password is like a spare key that is the same for every house the builder ever made. Find it once and you can open them all.

• The 42 sit in 37 advisories.
• One GE advisory adds 23 more. CISA describes them as "default or hard-coded credentials" but files them under a broader label. With those it is 65 (12.1%).
• Of the 93 critical flaws, 43 (46.2%) are authentication flaws.
• The FDA's guidance asks makers to test for credentials that are "'hardcoded,' default, easily guessed, and easily compromised."

Which weakness labels does CISA use most?

Improper Authentication (CWE-287) is the most used single label, on 45 of the 536 vulnerabilities (8.4%). A CWE is a label for the kind of mistake behind a flaw.

Weakness labels on 10 or more vulnerabilities

Source: The PenTest Index analysis of weakness labels printed in CISA ICS medical advisories; names from MITRE CWE. Checked October 9, 2026 (UTC).

A caution for anyone charting these. 167 of the 536 (31.2%) carry a label MITRE discourages or prohibits for mapping a specific flaw. Reasons include labels that are too broad, categories rather than weaknesses, and labels often confused with an attack's impact. That is why this page leads with groups.

Which medical devices and makers have the most vulnerabilities?

Imaging software has the most advisories: 43 of 192 (22.4%). Infusion pumps, medication and supply systems come next with 26, and patient monitors have 25. More advisories does not mean more danger.

A long list can mean more people are looking. It can also mean the maker reports its own flaws.

Which device types are named most?

Imaging machines have the highest share of high or critical flaws, at 76.6% (36 of 47). Heart devices (22.9%) and diabetes devices (19.4%) have the lowest.

CISA medical advisories and severity by device group

Source: The PenTest Index analysis of CISA ICS medical advisories, October 8, 2026. The ten groups are our own sorting, and every row is in the data file so you can sort them your way. A vulnerability listed in two groups is counted in each. We use the same selected score for that CVE in both groups. For CVE-2017-0143, the imaging-software row uses the 8.1 score in Baxter's ICSMA-20-170-01; Philips' ICSMA-18-058-02 gives only a range. Using only scores printed in imaging advisories would give 50 high or critical flaws out of 93 scored (53.8%).

Two notes keep this table fair.

• One GE advisory holds 23 of the 47 imaging machine flaws, all scored 9.8. Without it, imaging machines are 13 of 24 (54.2%).
• Heart and diabetes devices have more limited attack paths in many of these scores. For 30 of the 35 heart-device flaws and 19 of the 31 diabetes-device flaws, the attacker needs adjacent-network or physical access. Attack path is one part of the score; this table does not prove why the groups differ.

Which makers are named most?

Four makers are named in almost half of all advisories. Philips, BD, Medtronic and Baxter account for 88 of 192 (45.8%). This is not a ranking of who is worst.

The eight makers named in the most CISA medical advisories

Source: The PenTest Index analysis of CISA ICS medical advisories, October 8, 2026. Each advisory is counted under one maker. Maker names are tidied by us.

• There are 74 maker, vendor or software-project names in all, and 53 appear once. This includes software projects such as pydicom; it is not a count of 74 device manufacturers.
• Hillrom is counted on its own (3 advisories). Baxter completed its purchase of Hillrom in December 2021.
• BD is the only party credited in 16 of its 23 advisories. A maker that reports its own flaws will have a longer list.

Who finds or reports medical device vulnerabilities?

Mostly outside or anonymous reporters. Of the 192 advisories, 143 (74.5%) credit an outside or anonymous reporter, such as a security researcher or a hospital. The other 49 (25.5%) credit only the maker. An anonymous report does not tell us the reporter's affiliation.

In the versions we checked, 163 of 192 advisories (84.9%) used one of two narrower statements: no known public exploits, or no known public exploitation. Those statements do not mean the same thing.

Exploit wording on the advisory pages checked

Source: The PenTest Index reading of the exploit statement in each CISA ICS medical advisory, checked October 9, 2026 (UTC). The analysis cutoff is October 8, 2026. "Public exploit code" means code that could be used in an attack was public. It does not mean the device was attacked.

How did the list change after 2023?

The list looks different now. In 2016 to 2022, the four big makers were named in 57.7% of advisories and imaging software was 10.0%. In 2023 to 2026, the four makers fell to 21.0% and imaging software rose to 48.4%.

Before and after 2023

Source: The PenTest Index analysis of CISA ICS medical advisories, October 8, 2026. Device groups are our own sorting.

Of the 32 maker, vendor or project names in 2025 and 2026, 23 did not appear in the 2016–2024 medical-series advisories. The data shows the change. It does not explain it.

Which FDA cybersecurity alerts match a CISA advisory?

Eleven of the FDA's 18 cybersecurity safety notices match a CISA medical advisory in this dataset. Three more match a CISA advisory outside the medical series, three match broader CISA alerts, and one is general advice.

The FDA's table runs from June 13, 2013 to January 30, 2025. Nothing newer was in it when checked on October 9, 2026 (UTC).

FDA cybersecurity safety communications matched to CISA advisories

Source: FDA, Cybersecurity Safety Communications and Other Alerts, checked October 9, 2026 (UTC); matched by The PenTest Index to CISA advisories and CISA's published advisory files. "CVEs" counts ID numbers in the CISA record. It does not count attacks, patients or devices. The FDA's URGENT/11 press release links the medical advisory; the same flaws also appeared in ICSA-19-211-01. The Log4j and QNX counts cover the specific linked alerts, not every flaw in those software families.

Three things a writer should know before citing the FDA's table.

• It is not a full history. The FDA sent a letter about Illumina's Universal Copy Service on April 27, 2023. That letter is not in the table. CISA's matching advisory, ICSMA-23-117-01, is in our dataset.
• Advice changes after the first warning. The FDA updated its Contec notice on July 2, 2025 to say a patch exists. The FDA says the patch "fully removes networking functionality" and that patients and caregivers "should not install the software patch as the installation requires specialized expertise." Older text lower on the same page still says no patch is available.
• Others have studied these 18. A March 2026 paper in Frontiers in Digital Health analyzed the same notices. Our part is the match to CISA's records.

How do I look up a medical device, maker or CVE?

Search the table below by maker, product or CVE number. It holds all 192 advisories, and each row links to CISA's own page. No match means this list has no record. It does not mean a product is safe.

Search by maker, product or CVEMakerAllAbbottAliveCorAnimasApollo PharmacyB. BraunBaxterBDBeaconMedaesBiosense WebsterBIOTRONIKBMC Medical / 3B MedicalBoston ScientificBPL Medical TechnologiesCapsule TechnologiesCarestreamChange HealthcareContec HealthDario HealthDrägerEppendorfEthicon Endo-SurgeryFlow NeuroscienceFourth FrontierFresenius KabiFujifilmGE HealthCareGrassroots DICOMHamilton MedicalHillromi-SENSIlluminaINFINITT HealthcareInnokas Yhtymä OyInsuletInterpeak (IPnet; multiple RTOS vendors)LifePoint InformaticsMedixantMedtronicMicroDicomMirion MedicalNatus MedicalNextGen HealthcareNIHON KOHDENOFFISOpen Health Imaging FoundationOpenClinic GA (open source)OrthancOssurOxford Nanopore TechnologiesPanoramicPhilipsPixmeoPulsettopydicomQardioQualcomm LifeQuanovate Tech (Mira)Roche DiagnosticsSantesoftSiemensSmiths MedicalSoftnetaSOOIL DevelopmentsSpacelabsSt. Jude MedicalStrykerSwisslog HealthcareThermo FisherVarex ImagingVertikal SystemsVyaire MedicalWHILLYpsomedZOLLDevice groupAllBreathing, sleep and anesthesiaDiabetes devices and appsHeart implants, programmers and home monitorsHospital software, apps and device connectivityImaging machinesImaging software (PACS and DICOM)Infusion pumps, medication and supply systemsLab and diagnostic instrumentsPatient monitors, ECG and vital signsSurgery, therapy, shared parts and otherYearAll20162017201820192020202120222023202420252026Top severityAllCriticalHighMediumLowNo single scoreKnown exploited CVEAllyesnoCreditedAllDevice maker onlyOutside or anonymous reporter creditedCISA exploit statementAllNo known public exploitation reported (newer wording)No known public exploits (older wording)Nothing statedOther wordingPublic exploit code said to be availableSort resultsNewest advisory firstMakerCVE countTop CVSS v3 score

Copy filtered linkClear filters

Showing 192 of 192 advisories

Source: The CISA advisory source URL and check date appear in each row.

Where do the numbers everyone quotes come from?

Mostly from security companies, not from the government. The best-known figure, "6.2 vulnerabilities per medical device", traces to a February 7, 2018 vendor blog post with no sample, dates or method. The FBI printed it in 2022 without naming the research firm.

The numbers people repeat, and where each one started

Source: each linked original, checked October 9, 2026 (UTC). Quoted words are copied from those sources; the rest of the table explains their scope.

Here is where "6.2" appears. A 2018 vendor blog post. A 2021 trade article that names Sensato. A 2022 FBI notice that calls it "a research report in 2021" by "a cybersecurity firm" and names no one. A 2023 GAO report that cites the FBI. The FBI does not identify its 2021 source, so we cannot prove that step in the chain. The same vendor figure now reads like a government fact.

Why do the numbers disagree so much?

Because they answer three different questions. One counts flaws, one counts devices, and one counts attacks.

Three questions, three kinds of answer

Source: The PenTest Index analysis and the Claroty press release linked above, checked October 9, 2026 (UTC).

Units trip people up most. Claroty reported its riskiest mix (a known exploited flaw tied to ransomware, plus an insecure internet connection) in 1% of devices but 89% of organizations. Both are right. One hospital can own thousands of devices, and it takes only one to count the hospital.

The same report shows how much device type matters. It puts that riskiest mix at 8% of imaging systems and 0.5% of patient devices in its sample.

What do CVE, KEV and CVSS mean?

They do three different jobs. A CVE names a flaw, KEV says a flaw has been used in real attacks, and CVSS scores how bad a flaw could be.

Terms that get mixed up

Source: definitions from CISA, FIRST and MITRE, put in plain words by The PenTest Index.

Does the FDA require penetration testing for medical devices?

The law does not name penetration testing, but the FDA's guidance recommends it. Section 524B applies to qualifying FDA applications or submissions for "cyber devices" filed on or after March 29, 2023. It requires a security plan, a patching process and a list of software parts. The FDA's guidance, dated February 3, 2026, lists five things a penetration test report should include.

A penetration test is when trusted experts try to break into a product the way a real attacker would, so the maker can fix what they find.

The law and the guidance, side by side

Source: FDA, Cybersecurity and the premarket cybersecurity guidance, checked October 9, 2026 (UTC). A "cyber device" has software and technology validated, installed or authorized by its sponsor, can connect to the internet, and could be open to cyber threats. The sponsor is the person or company making the FDA submission. Read the FDA's exact definition and submission rules. The guidance also covers devices with software, firmware or programmable logic that do not connect to a network.

The FDA says penetration test reports "should be provided and include the following elements":

1. Independence and technical expertise of testers
2. Scope of testing
3. Duration of testing
4. Testing methods employed
5. Test results, findings, and observations

Three more lines from the same guidance tie back to the data on this page.

• On exploited flaws: vulnerabilities in CISA's Known Exploited Vulnerabilities Catalog "should be designed out of the device, as they are already being exploited." All 12 in this dataset came in through outside software.
• On passwords: makers should test for credentials that are "'hardcoded,' default, easily guessed, and easily compromised." Forty-two flaws here concern hard-coded or default passwords and keys.
• On timing: after release, cybersecurity testing "should be performed at regular intervals commensurate with the risk (e.g., annually)." Annual testing is an example based on risk, not a rule requiring every device to have a yearly penetration test.

One thing may confuse a careful reader. On October 9, 2026 (UTC), the FDA's own cybersecurity page still showed the June 27, 2025 date for this guidance. The PDF's cover says February 3, 2026.

If your device also has a web app or API, you can compare web and API penetration testing offers by the prices and scope they publish.

What should you do with a medical device vulnerability notice?

Match the notice to your exact product and software version first, then follow the maker's current instructions. Of 192 advisories, 68 show a later revision date than their number, so the first version you read may not be the last word.

If you make a device

• Check your software parts list against CISA's exploited list. All 12 known exploited flaws here came in through outside software.
• Test how the product checks identities. More than one in four flaws here is an authentication flaw.
• Plan for a penetration test report that covers the FDA's five items.

If you run devices in a hospital

• Look up your devices in the table above, then open CISA's page for each.
• Ask each maker for its software parts list and its patch plan.
• Work with the team that owns clinical safety before you change a device that is in use.

If you use one of these devices yourself

• This page is not medical advice.
• Follow your device maker's patch instructions. Do not try unofficial fixes you find online. Ask your care team if you are unsure. The FDA's plain-language guide is a good place to start.

Say you are a small device maker with a connected glucose meter and an FDA filing next spring. You would want a tester who can show independence, a clear scope, how long the test ran, the methods, and the findings. Those are the FDA's five items.

Getting ready to buy a penetration test for a device? Find My PenTest Match is our free scope checklist. It helps you write down what needs testing before you contact anyone.

How did we build this?

We checked all 192 CISA medical advisory pages on October 9, 2026 (UTC), including each CVE, weakness label, score, vector and credit line. We matched each CVE to CISA's Known Exploited Vulnerabilities catalog (version 2026.10.08, 1,739 entries). The analysis cutoff is October 8, 2026. The core counts come from the two advisory and CVE files. FDA matches, the linked maker table and quoted figures have their own source files.

Where the data came from

• CISA's ICS medical advisory pages: 191 in the medical section, plus one with a medical number that CISA files with its other advisories (ICSMA-18-137-02).
• CISA's Known Exploited Vulnerabilities catalog.
• MITRE's CWE list, version 4.20, for weakness names and groups.
• FDA notices, recall records and guidance for the FDA sections; the original producer of each figure in the source-check table.
• Johnson & Johnson's original Carto 3 PDF for the separate linked-table check.

The rules we used

1. One CVE is one vulnerability, however many advisories list it.
2. Scores and weakness labels are the ones printed on CISA's pages. Where a CVE has a score in one advisory and only a range in another, we use the score.
3. Severity bands follow FIRST's CVSS v3.1 scale.
4. "Authentication flaw" means CWE-287 and everything MITRE files under it. The CSV field is authentication_flaw; it includes certificate and device-identity checks, not just logins. "Hard-coded or default passwords and keys" means CWE-798, CWE-259, CWE-321, CWE-1392 and CWE-1394.
5. Each advisory gets one maker, vendor or project label and one of ten device groups. Both are our own sorting. The infusion and dispensing group includes medication and supply cabinets.
6. "Year" is the year in CISA's advisory number.
7. We count CVEs in the vulnerability descriptions, not stray IDs in navigation or unrelated references. In the Contec advisory, the credit paragraph has a typo, CVE-2025-1024. The vulnerability section, update log and researcher's disclosure identify CVE-2025-1204. We count the correct ID once.
8. We count each CVE once overall. For a CVE listed more than once, a single published score takes priority over a range-only row. For CVE-2017-0143, this uses Baxter's 8.1 score in every group where that CVE appears. The per-advisory table keeps each page's own scores. All other repeated CVEs have matching values.
9. We count outside or anonymous reporting credit together. It is a count of the credit lines, not proof that each reporter is independent or that a penetration test found the flaw.
10. Percentages use the denominator printed beside them. They are rounded separately, half up, to one decimal.
11. For the FDA recall count, retrieve all records matching reason_for_recall:cybersecurity, then count distinct res_event_number values. The 60 returned product records form 17 events.
12. For the Carto 3 check, take the explicit CVE IDs on PDF pages 8–13, remove duplicates, join them to the core CVE IDs, then match that union to the same KEV catalog. The 183 IDs include 21 already counted: 536 + 183 − 21 = 698; 20 / 698 = 2.9%.

How we checked ourselves

• We independently recomputed the submitted data and rechecked all 192 source pages. The 551 CVE listings and 536 distinct IDs matched.
• We compared the inventory with CISA's medical advisory index and official advisory data repository. We used CISA's web pages when the versions disagreed.
• We worked out all 526 populated score-and-vector rows again. All matched. The other 25 listings have only ranges; one repeated CVE has a single score in another advisory, leaving 24 distinct CVEs without a single score.
• Our yearly advisory counts for 2016 to 2024 match the ones MedCrypt published in 2025. MedCrypt counted these advisories before we did.

Where CISA's data files and its web pages disagree, we used the web pages. For example, the repository inventory lacks four advisories, and its ICSMA-21-187-01 file omits CVE-2021-39369, which the page prints. Some scores, vectors and weakness labels also differ. We do not mix in the repository's extra single scores for page rows that give only ranges.

What does this data show, and what does it not show?

It shows what CISA has published about medical device flaws. It does not show how many devices in use are at risk today.

• It counts notices, not danger. A maker that reports its own flaws gets a longer list.
• "Known exploited" is narrow. Here it means a match to CISA's catalog. The 12 matches do not tell us that the other 524 have never been used, or that these attacks took place on medical devices.
• It does not track fixes. A flaw named in 2018 may be long patched, or not.
• Scores rate software, not patient harm.
• Groups and maker names are ours. Sort the file differently and the device and maker tables will shift a little. The totals will not.
• It cannot explain trends. We can see imaging software rise. We cannot see why.
• It is U.S. government data. Flaws reported only to other countries' agencies, or never reported, are not here. CISA's medical category includes hospital software and shared components; it is not a list of FDA-cleared devices or a U.S. device sample.

You can read how The PenTest Index works and how we make money.

How do I cite this page?

Credit The PenTest Index for the counts and the matching, and keep CISA named as the source of the advisories. Keep the date with the number: 536 directly named CVEs and 12 catalog matches use the October 8, 2026 cutoff. Sources were checked October 9, 2026 (UTC).

Copy this citation:

The PenTest Index. "Medical Device Vulnerabilities: 536 CVEs, 12 on CISA's Exploited List." Data as of October 8, 2026; verified October 9, 2026 (UTC). https://thepentestindex.com/research/medical-device-vulnerabilities/

Or copy the finding in one sentence:

CISA's medical advisory pages directly name 536 distinct medical device vulnerabilities in 192 notices since 2016; 12 (2.2%) are on CISA's exploited list, all Microsoft or Apache flaws, according to The PenTest Index's October 2026 analysis of CISA data.

Can I reuse the tables, charts and data?

Yes. You may reuse our original counts, charts and device-group sorting; credit The PenTest Index by name. Keep CISA named as the source of its advisories and catalog; CISA releases the catalog under CC0. MITRE's CWE material and any manufacturer or vendor text keep their own attribution and terms. This permission covers our contribution, not rights in underlying source material.

Where can I download the data?

Every row behind this page is free to download as CSV, and each row carries its source link and check date.

• medical-device-advisories.csv: all 192 advisories, with maker, device group, CVE count, top score, known exploited count, credit and CISA's exploit statement.
• medical-device-cves.csv: all 551 CVE listings (536 different CVEs), with weakness label and group, score, vector and known-exploited status.
• fda-notices-matched-to-cisa.csv: the FDA's 18 notices plus the 2023 Illumina letter, matched to CISA records.
• quoted-numbers-source-check.csv: the repeated numbers, with each original quote and link.
• findings-with-math.csv: main findings, with counts, denominators and the math.
• carto3-referenced-cves.csv: 183 CVE IDs from Johnson & Johnson's linked Carto 3 table, with PDF pages and overlap and KEV checks.
• The four charts as PNG files, each with its source line inside the picture: exploited-list matches, advisories by year, weakness groups, and attack paths.

Source: The CISA advisory source URL and check date appear in each row; KEV fields use the catalog version stated in the data.

Questions people ask

What is a medical device vulnerability?

A medical device vulnerability is a security flaw in a medical device or its software. Someone could use it to break in, steal data, or change how the device works. CISA has named 536 of them in 192 advisories since 2016.

How many medical device vulnerabilities are there?

CISA's medical advisory pages directly name 536 in 192 advisories, from March 29, 2016 to October 8, 2026. Two early advisories also report 1,418 and 460 flaws in outside software without naming each one. Their sum is 1,878, but unknown overlap means we cannot add it to the 536 distinct CVEs.

What is the most common medical device vulnerability?

Access control. Of the 536 vulnerabilities, 210 (39.2%) are access control flaws and 152 (28.4%) are authentication flaws: weak checks of a claimed identity. Forty-two are hard-coded or default passwords and keys.

Does a known vulnerability mean a medical device was hacked?

No. A vulnerability is a flaw that could be used in an attack. Only 12 of the 536 are on CISA's list of flaws used in real attacks, and even that list does not say which devices were hit.

Which medical devices have the most security advisories?

Imaging software has the most: 43 of 192 advisories (22.4%). Infusion pumps, medication and supply systems have 26, and patient monitors have 25. More advisories does not mean more danger.

Is there a medical device vulnerability database?

CISA posts each medical advisory on its website, and the FDA lists 18 cybersecurity safety notices. This page puts all 192 CISA advisories in one free file, with each flaw's score, weakness type and known-exploited status.

What percentage of medical devices are vulnerable?

These sources do not give one current rate for all medical devices. Claroty reported in March 2025 that 9% of the connected medical devices it analyzed carried a known exploited flaw. Always keep the company, the year and the unit with the number.

Where does "6.2 vulnerabilities per medical device" come from?

The earliest source we found is a February 7, 2018 blog post by a security vendor, Sensato. It gives no sample, dates or method. The FBI repeated the figure in 2022, and the GAO repeated the FBI in 2023.

Should a patient stop using a device named in an advisory?

A table cannot decide that. Talk with your care team and follow the maker's current instructions. The FDA says that for the 18 cases it lists, it is not aware of any patient injuries or deaths tied to cybersecurity incidents.

Sources

Sources were checked on October 9, 2026 (UTC). The analysis cutoff is October 8, 2026; the KEV catalog version is 2026.10.08. Historical figures keep the dates of their original reports.

1. CISA, ICS advisories and ICS medical advisories (192 advisory pages). https://www.cisa.gov/news-events/ics-advisories
2. CISA, Known Exploited Vulnerabilities catalog, version 2026.10.08. https://www.cisa.gov/known-exploited-vulnerabilities-catalog
3. CISA, how vulnerabilities are added to the catalog. https://www.cisa.gov/known-exploited-vulnerabilities-catalog/reducing-significant-risk-known-exploited-vulnerabilities
4. CISA, advisory data files (cross-check). https://github.com/cisagov/CSAF
5. CISA, catalog data and CC0 license. https://github.com/cisagov/kev-data
6. Johnson & Johnson, Carto 3 operating-system patch advisory, April 2018. https://www.productsecurity.jnj.com/sites/default/files/2023-10/CARTO3v4-Advisory-040918.pdf
7. MITRE, CWE view 1400. https://cwe.mitre.org/data/definitions/1400.html
8. FIRST, CVSS v3.1 specification and user guide. https://www.first.org/cvss/v3.1/specification-document
9. FDA, Cybersecurity (safety communications table). https://www.fda.gov/medical-devices/digital-health-center-excellence/cybersecurity
10. FDA, Cybersecurity in Medical Devices: Quality Management System Considerations and Content of Premarket Submissions, February 3, 2026. https://www.fda.gov/media/119933/download
11. FDA, Contec and Epsimed patient monitors safety communication, updated July 2, 2025. https://www.fda.gov/medical-devices/safety-communications/cybersecurity-vulnerabilities-certain-patient-monitors-contec-and-epsimed-fda-safety-communication
12. FDA, Illumina Universal Copy Service letter, April 27, 2023. https://www.fda.gov/medical-devices/letters-health-care-providers/illumina-cybersecurity-vulnerability-affecting-universal-copy-service-software-may-present-risks
13. FDA, device recall data (openFDA), data updated October 8, 2026. https://api.fda.gov/device/recall.json?search=reason_for_recall:cybersecurity&limit=100
14. FDA, Medical Device Cybersecurity: What You Need to Know. https://www.fda.gov/consumers/consumer-updates/medical-device-cybersecurity-what-you-need-know
15. GAO, GAO-24-106683, December 21, 2023. https://www.gao.gov/assets/gao-24-106683.pdf
16. FBI, Private Industry Notification 20220912-001, September 12, 2022. https://www.ic3.gov/CSA/2022/220912.pdf
17. Sensato, blog post, February 7, 2018. https://www.sensato.co/post/endless-terrifying-possibilities-call-for-a-good-medical-device-cop
18. MedTech Dive, December 1, 2021. https://www.medtechdive.com/news/cybersecurity-medical-devices-hospital-divide-fda/609252/
19. Cynerio, press release, January 19, 2022 (archived copy). https://web.archive.org/web/20220527214113/https://www.cynerio.com/blog/cynerio-research-finds-critical-medical-device-risks-continue-to-threaten-hospital-security-and-patient-safety
20. Unit 42, Palo Alto Networks, March 2, 2022. https://unit42.paloaltonetworks.com/infusion-pump-vulnerabilities/
21. Claroty, press release, March 26, 2025. https://claroty.com/press-releases/new-research-from-clarotys-team82-highlights-riskiest-medical-device-exposures-in-healthcare-environments
22. Claroty, State of CPS Security: Healthcare Exposures 2025. https://web-assets.claroty.com/resource-downloads/state-of-cps-security-healthcare-2025.pdf
23. Health-ISAC, Finite State and Securin, 2023 State of Cybersecurity for Medical Devices and Healthcare Systems. https://health-isac.org/wp-content/uploads/11883-StateMedSecurityReport_v6.pdf
24. MedCrypt, ICS-CERT medical device advisory trends, April 2, 2025. https://www.medcrypt.com/cybersecurity-whitepapers/whitepapers-ics-cert-2024-medical-device-cybersecurity-trends
25. Menon, Frontiers in Digital Health, March 2026. https://www.frontiersin.org/journals/digital-health/articles/10.3389/fdgth.2026.1701551/full
26. FDA, Cybersecurity in Medical Devices Frequently Asked Questions. https://www.fda.gov/medical-devices/digital-health-center-excellence/cybersecurity-medical-devices-frequently-asked-questions-faqs
27. Baxter, acquisition of Hillrom completed December 13, 2021; company release filed with the SEC. https://www.sec.gov/Archives/edgar/data/10456/000162828021024946/bax-20211213pressreleasexe.htm
28. FDA, URGENT/11 press release, October 1, 2019. https://www.fda.gov/news-events/press-announcements/fda-informs-patients-providers-and-manufacturers-about-potential-cybersecurity-vulnerabilities
29. MITRE, CWE version 4.20 data. https://cwe.mitre.org/data/xml/cwec_v4.20.xml.zip

The PenTest Index Research is the research and reference section of thepentestindex.com.