Where connected medical systems and clinical AI actually fail — not in theory, but in the moment when the decision still has to be made. Built from 28 years of clinical reality and lived device dependency.
Explore real healthcare failure scenarios, make the call in Decision Mode, and see where security frameworks stop short of patient safety.
At 2am in the ER, I'm not waiting on a dashboard. Healthcare doesn't pause for data validation. Our security frameworks need to stop assuming it does.
Most healthcare cybersecurity focuses on protecting data. This lab focuses on what happens when the technology keeping a person alive fails — in real time, under real conditions, with real consequences that no framework fully accounts for.
I never left healthcare. I learned to defend it. I teach it, I live it, I build it, and I'm learning how attackers see it — so I can protect patients better than anyone who only knows one side.
Every scenario carries an evidence label, a failure chain, a patient-side reading on the five PSTM lenses, a control and framework analysis, and its sources. Tap a scenario to open it.
Battery indicator reports full charge. Device loses power within minutes. No backup source. Power infrastructure unstable.
Evidence basisBuilt from the author’s own device dependency. The failure chain is lived experience, not a published incident.
AI triage scores patient low acuity. Clinician overrides based on observation. System flags the override as anomalous.
Evidence basisModeled on how triage AI and override logging interact. No specific product or incident is described.
Networked pump receives unauthorized parameter update during active medication delivery. Nurse managing four other patients simultaneously.
Evidence basisNetworked infusion pump vulnerabilities are documented by FDA (Hospira Symbiq, 2015) and CISA advisories. This attack chain is a composite, not a single reported event.
Before attackers touch a network, they watch it. What they find in 30 minutes would surprise most security teams.
Evidence basisHHS HC3 reported help-desk social-engineering attacks against health organizations in 2024. The hospital and caller here are composites.
Your Apple Watch says you're fine. The clinical system believes it. The coach clears you to play. You are not fine. Two perspectives — same failure, different stakes.
Evidence basisNo public attack on a specific wearable is described. Modeled on how consumer wearable data enters clinical and clearance decisions.
This is the scenario that bridges healthcare and sports technology. Same device. Same biometric data. Two entirely different contexts — both with life-threatening stakes. The gap between consumer wearable and clinical medical device is where attackers operate, and where the frameworks reviewed here draw no clear line.
The EHR is locked. The patient is crashing. 90 seconds. No medication history. This is not a drill.
Evidence basisEHR downtime, paper charting and diversion are documented in real attacks, including Ascension (2024) and Luminis Health (2026). The patient case is a composite.
Connected to the internet. Can receive software updates remotely. The patient depending on it is also the analyst who built this lab.
Evidence basisNo public CVE for the AirSense 11 was found as of September 2026. This models the remote-update path as an attack surface.
A patient with major neurocognitive disorder (DSM-5; formerly called dementia) living at home relies on a connected ecosystem of smart sensors, GPS tracking, automated medication dispensers, and telehealth platforms. Each device is a lifeline. Each connection is a potential failure point. The patient cannot self-advocate when the system fails.
Evidence basisModeled on a real care pattern: connected home monitoring for a family member living alone. No specific product or incident.
A compromised continuous glucose monitor transmits falsely normal readings. An endocrinologist makes an insulin dosing decision on manipulated data. The patient never knew the signal was wrong.
Evidence basisNo public exploit against the Dexcom G7 is known. This models where integrity can fail between sensor and screen.
Select a scenario — read the situation — make the call.
Real incidents. Real patient impact. Real framework gaps. Attacks on hospitals are attacks on human life — not just data. A few of the largest below — the full searchable tracker has the complete dataset.
The same scenarios, decisions and patient-side method, run against your environment, with your clinicians and your security team in the room.
A public research environment for healthcare cyber risk: where connected medical devices, clinical AI and hospital systems fail, and what that failure does to a patient.
From clinical practice, public incident reporting, regulatory guidance, and daily device dependency: pulmonary hypertension managed with a portable oxygen concentrator and a CPAP connected to the internet. Every scenario is labeled as a documented incident, a documented vulnerability, a threat model, or an author-developed scenario, and lists its sources.
Patient-Side Threat Modeling (PSTM) reads each failure through five lenses: physiological dependency, latency to harm, bedside detectability, workaround pressure and downtime survivability. Response follows the Digital First Responder sequence: Triage, Stabilize, Treat, Recover.
As source material for tabletop exercises, downtime drills, clinical AI reviews and training. See how to bring it in →
Created by Chaunda C. Dallas, MSIT, Healthcare Cybersecurity Strategist. chaundacdallas.com →