Report 251 · Luxury Home Security
What physical access means on a camera advisory
In September CISA published an advisory on a CareCam Pro IP camera with a hard-coded password in its bootloader. It scored Medium under the older scoring system, because the attacker has to touch the camera. That discount rests on an assumption: few people can get their hands on the device. For a camera mounted at the edge of a property, getting close is exactly what it is installed to allow.
The advisory is ICSA-26-251-01, released 8 September 2026. It covers one model, the ANJIA AJL33PC0801, sold under the CareCam Pro name, and one flaw, CVE-2026-85083. CISA's description: the camera "uses a hard-coded credential for bootloader authentication. An attacker with physical access to the device may leverage this weakness to gain privileged bootloader access, allowing unauthorized modification of firmware and system configuration and potentially resulting in complete device compromise."
The bootloader is the small program that runs before the camera's operating system and decides what software loads. A password baked into it is the same on every unit and cannot be changed by the owner. Anyone who knows it and can reach the right connection on the device can, by CISA's account, rewrite what the camera runs.
The affected firmware string tells its own story: a bootloader version from 2010, firmware compiled on 26 August 2020. Under Mitigation, the advisory says only: "CareCam has not responded to CISA's attempts for coordination. Users are encouraged to reach out to CareCam." There is no patch to install. CISA adds that no known public exploitation has been reported and that the "vulnerability is not exploitable remotely."
Two scores, one flaw
CISA lists the flaw under both versions of the Common Vulnerability Scoring System. Under CVSS v3.1 it is 6.8, MEDIUM. Under CVSS v4.0 it is 7.0, HIGH. Both vectors start the same way: AV:P, attack vector physical. The impact metrics are all at the top of the scale (confidentiality, integrity and availability all High). What holds the number down is the attack vector.
I explained how to read a vector string, instead of the headline score, in Report 131. This one is worth reading letter by letter, because a single metric is doing almost all the work.
What "physical" assumes
The CVSS v3.1 specification from FIRST defines Physical as an attack that "requires the attacker to physically touch or manipulate the vulnerable component," and notes the contact "may be brief" (its example is the "evil maid" attack, where someone alone with a laptop for a few minutes tampers with it). The v4.0 specification uses the same definition.
The reason physical attacks score lower is stated plainly in the v3.1 spec: the score "will be larger the more remote (logically, and physically) an attacker can be," on the assumption "that the number of potential attackers for a vulnerability that could be exploited from across a network is larger than the number of potential attackers that could exploit a vulnerability requiring physical access to a device."
That assumption is sound for a server in a locked rack or a laptop in a bag. The smaller group of people who can touch it is the whole argument. Coming from counter-IED and electronic warfare, I treat "who can reach it, and when?" as the first question about a device, not a footnote. A camera mounted on a gate post, a driveway pillar or a back wall facing a trail is outside the house by design. You put it there because people approach that spot. On a lot of estates, the person who can get within arm's reach of a perimeter camera does not have to break in to do it. They need a ladder, a few minutes and a reason to be near the fence line.
None of this makes CISA's score wrong. A base score is meant to be generic. The specification says so: base metrics "are common to all environments," and "Consumers of CVSS should supplement the Base Score with Temporal and Environmental Scores specific to their use of the vulnerable product." The v4.0 document adds that product maintainers and databases "typically provide only the Base Scores." The score you read in an advisory was never meant to describe your property. Adjusting it for where the device is mounted is your job, and almost no homeowner does it.
What a rewritten camera means
CISA's words are "unauthorized modification of firmware and system configuration" and "complete device compromise." My reading, since the advisory does not go further: once someone controls the firmware, you can no longer trust what the camera shows you, what it records or where it sends it. A camera that has been tampered with can keep looking perfectly normal in your app. That is what separates this from cutting a wire. A dead camera announces itself. A rewritten one may not.
The second point is where the camera sits on your network. A compromised device inside your home network is a foothold next to everything else on it. That is why CISA's standard recommended practices in the advisory begin with minimizing network exposure and isolating devices behind firewalls. Those practices are generic text CISA attaches to many advisories, but they fit this case well.
What I would do with an outdoor camera like this
I help design the AI security systems for a veteran-owned (SDVOSB) home-security company run by fellow veterans. I do not own that company and earn nothing from this link. Full policy here. From that seat, this is the checklist I would run for any camera whose advisory says "physical access required":
- Find out what you own. Check model numbers against CISA advisories, not just the brand on the app. This advisory names one specific model.
- Walk the perimeter as an outsider. For each outdoor camera, ask whether someone could reach it without crossing a fence, triggering a sensor or being seen by another camera. If yes, treat "physical" as close to "remote" for that device.
- Make tampering loud. Mount it higher, in a tamper-resistant housing, with its cable protected. Have the recorder alert on video loss or a sudden change in view, and have a second camera covering the first.
- Contain it on the network. Put cameras on their own network segment so a rewritten one cannot reach your other devices.
- Plan to replace it. The vendor has not answered CISA, and a hard-coded bootloader password is a design choice, not a bug you can configure away. As I wrote in Report 60, when no patch is coming, the fix is new hardware.
The general lesson
"Not exploitable remotely" is true and useful. It means no one can do this from another continent. It does not mean the device is out of reach. When you read an advisory, find the AV letter in the vector and ask one question: for this device, where it is mounted, who can actually touch it? For cameras at the edge of a property, the honest answer is often "more people than the score assumes."
Sources
- Cybersecurity and Infrastructure Security Agency, "CareCam Pro IP Cameras," ICS Advisory ICSA-26-251-01, released 8 September 2026 (initial publication, no revisions as of reading). (Primary source, full advisory read. Source for the model, CVE-2026-85083, CWE-798, the affected firmware/bootloader string, the vulnerability description and the vendor-response and "not exploitable remotely" statements quoted verbatim, the CVSS v3.1 6.8 MEDIUM and v4.0 7.0 HIGH scores and vectors, no known public exploitation, and the recommended practices. CISA lists the sector as Commercial Facilities and deployment as worldwide; it does not say how many units are in homes.)
- FIRST, "Common Vulnerability Scoring System v3.1: Specification Document." (Primary source, read. Source for the Physical attack-vector definition, the "more remote" scoring rationale and the statements on Base versus Environmental scores, quoted verbatim.)
- FIRST, "Common Vulnerability Scoring System v4.0: Specification Document." (Primary source, read. Source for the v4.0 Physical definition and the statement that providers "typically provide only the Base Scores.")
- Prior reporting in this publication: Report 131 (reading a CVSS vector); Report 60 (when a camera has no patch). (Context only.)