Search
Find a vulnerability
Search criteria
1 vulnerability found for LineEds by Circutor
GCVE-1988-2026-0467
Vulnerability from gna-1988 – Published: 2026-11-02 07:30 – Updated: 2026-11-02 07:30
VLAI
EPSS
VEX
Title
[0day-rubbish] Circutor LineEds 24.11.14-r0 unauthenticated pwrstudio events.xml shellExecute command injection (9.8)
Summary
0day Rubbish Research Team is publicly disclosing a vulnerability in the pwrstudio
daemon shipped in Circutor LineEds (Line Energy Data System) industrial energy gateway
firmware 24.11.14-r0, engine variant pss0. Circutor S.A. is in Spain; the appliance sits
between metering and power-quality instrumentation on one side and an energy-management
or SCADA back office on the other.
Type: operating-system command injection (CWE-78) reached without authentication
(CWE-306). Those two are the classes recorded in the research; the advisory's own
classification additionally maps CWE-285, because the permission gate fails open in two
independent ways, and CWE-20, because the events document is accepted with no schema
validation, as contributing weaknesses, with CWE-250/CWE-269 recorded as bearing on
remediation priority rather than on the defect.
Root cause: the daemon's event engine loads its configuration from events.xml, exposed at
/services/user/events.xml over the management HTTP interface. An event may carry a
shellExecute action whose command and parameter elements are concatenated by
XCEventActionShellExecute::Execute (FUN_00548414) with the wide-string format "%ls %ls"
stored at 0x7cc2d0, converted from wide to narrow, and passed to system() at call site
0x5484bc. Nothing between the socket and the sink neutralises shell metacharacters: no
quoting, no filtering, no allowlist. The format string is the whole reason the injection
works the way it does - two %ls conversions separated by a single space mean the sink
performs concatenation rather than argument separation, so command and parameter become one
shell word sequence with no quoting boundary between them; a payload needs only a statement
separator, not a quote escape. The defect is at that concatenation rather than in the XML
parser, which performs a pure element-to-field copy and is behaving correctly.
Exploitation is two HTTP requests with no delivery precondition of any kind. Step one
writes an event definition into /services/user/events.xml carrying an always-true
condition so it enters the live set on every evaluation pass, a deliberately invalid
command token so the first shell statement fails harmlessly, and a parameter of the form
semicolon, command, semicolon, hash - yielding a shell line such as
"x ;id >/tmp/circutor_eds_vuln_rce_marker;#". Step two requests the restart.xml route,
which makes the engine reload the document through XCEventDriver::LoadXml (FUN_0020acac),
build the action object, evaluate the condition and dispatch Execute. There is no race, no
memory-layout dependency, no file to pre-place on the target and no dependence on another
vulnerability, and because the intermediate state is a file the product itself persists,
the two steps can be separated in time arbitrarily. The binary is compiled with
CONFIG_NO_HASP, so no licence-dongle gate stands between a written event and the shell
action. The image is a non-PIE EXEC, so every offset quoted is a fixed file-and-memory
address reproducible against the same build with no relocation arithmetic.
Unauthenticated classification, and it does not rest on a configuration assumption. The
product implements authorization as a per-handler gate rather than as a dispatcher
middleware stage: the gate is one function, FUN_001a93cc, invoked with a permission
constant by whichever handler chooses to invoke it. A cross-reference census of that
function in the shipped binary records 197 call sites, all individual action handlers, all
inside the address range 0x1aexxx to 0x1bfxxx. Both handlers in this chain fall outside it:
the restart handler FUN_001e3554 at 0x1e3554 and the events write handler FUN_0023ace0 at
0x23ace0. The gate is therefore structurally incapable of authorizing either, because it is
never invoked on their paths. The restart handler's only guard is FUN_001a955c, a
three-instruction predicate testing that the request-context fields at +0x34 and +0x4c are
both non-null; those are the request and response object pointers installed
unconditionally by the dispatcher's own setup routine FUN_001add40, which the route entry
at 0x76a480 names at offset +0x0c, ahead of the handler at +0x10. It tests object presence,
not credentials. One honest tension is published rather than smoothed over: the events write
handler and its dispatcher contain no authorization call that we could find, yet the handler
is recorded as capable of both 401 and 200 outcomes and we could not determine what selects
the 401 branch - so a credential-required conditional score is carried alongside the
primary rating precisely to account for that unresolved path.
A second, configuration-independent defect is documented in the same advisory. The gate's
own logic returns success with no check of any kind when the permission constant is not
0x15, and returns success when enforcement is toggled off, so it fails open in two
independent ways for the 197 handlers that do call it. This chain does not depend on that
defect, and the advisory deliberately makes no claim that any specific other route is a
second path to command execution; it is reported as a separate finding for vendor review.
Two claims carried by the research record are deliberately NOT published as fact. First,
that security enforcement is off by default: no shipped artifact supports the default value
- nothing inside the image, no value read from the extracted root filesystem, not the
contents of the engine.xml package, and no vendor documentation of the factory setting. The
unauthenticated classification does not depend on it, because neither route in the chain
invokes the gate at all; the default matters only for the alternative reloadCfg.xml
trigger, whose reload action does call the gate with permission 0x15, and that variant is
treated as conditional. Second, that pwrstudio runs as root on a device: the observed
uid=0(root) is the privilege of the root-run emulation harness, and the device-side account
is inferred from the daemon's ownership of the restart, reload and upgrade routes, since no
init script or service unit was extracted. Neither caveat alters the impact metrics.
Scoring. Four readings were priced and all four are published, two of them as not adopted:
- PRIMARY, 9.8 Critical, CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
- CONDITIONAL, 8.8 High, CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H, applying where a
deployment rejects anonymous configuration writes - enforcement enabled, or the events
handler's unresolved 401 branch active - so that any valid operator credential suffices.
Nothing else in the chain changes.
- NOT ADOPTED, 10.0 Critical, CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H - the Scope
Changed reading a reviewer who counts the appliance OS as a realm distinct from the
management daemon would apply. We do not adopt it: the command executes inside the
authority the vulnerable component already holds and no sandbox, container,
virtualisation boundary or separate security realm is crossed. Published so the choice is
visible and so the higher number is not mistaken for our claim.
- NOT ADOPTED, 8.1 High, CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H - Attack Complexity
raised despite the absence of any attacker-unsatisfiable precondition.
A PR:H reading of 7.2 is not carried, because nothing indicates the configuration write or
the restart route demands an administrator rather than any authenticated principal.
On scoring accuracy: the research record carried 9.8 with that primary vector and
recomputing it under the CVSS 3.1 base algorithm yields 9.8, so the recorded number is
arithmetically correct. The reconciliation that matters is the grounding rather than the
arithmetic: the record derived PR:N from an asserted factory default for the security
toggle, for which no shipped artifact exists, whereas the advisory derives PR:N from the
handler call census and the two handler decompilations, which are artifact evidence. The
score is unchanged and the reason for it has been replaced with one that survives scrutiny.
We note this because several earlier advisories in this repository carried recorded scores
that did not compute for their recorded vectors.
Impact: arbitrary command execution on an industrial energy gateway with the daemon's
privileges. Confidentiality High - every measurement record, configuration item, credential,
certificate and key on the appliance becomes readable, including whatever it holds to
authenticate to the energy-management back office, and a compromised gateway is an ideal
passive collection point because it sees the traffic it forwards. Integrity High -
measurement data can be altered in transit or at rest before reaching the back office, so
forged consumption, generation or power-quality figures propagate into reporting,
settlement, alarm thresholds and control decisions while arriving through the legitimate
path and are hard to distinguish from real data; event configuration can be rewritten and
firmware pushed through the same management plane, since the upgrade route is served by the
same daemon. Availability High - the attacker controls the very restart and reload
primitives this chain uses, and can reboot on demand, stop measurement collection, drop
configuration or wedge the engine, producing a data outage in the metering path.
Persistence and lateral movement are High in practice on the inference below: a foothold
inside the OT segment on a device that security tooling normally treats as trusted
instrumentation.
Verification boundary, stated plainly. Confirmation used the shipped binary extracted from
the firmware package, run under qemu-arm-static USER-MODE emulation with gdb - no
full-system emulation, so no kernel, device drivers, network stack or reverse-proxy front
end were present. No physical device was used at any point: nothing was powered, connected,
flashed or queried. The daemon's HTTP stack could not be started under emulation, for a
specific reason: its serial initialisation thread loops on an open of a Freescale i.MX
private serial device returning EACCES, the device does not exist under qemu-user and no
device model is provided, so the engine-ready futex is never signalled and the worker thread
that would accept connections is never scheduled. A listening socket on port 1500 does come
up with an accept backlog, but nothing ever issues accept or poll against it, so no request
can be delivered and no response produced.
The real shipped sink function was therefore invoked directly under debugger control with
the field values the documented XML schema produces, using a return-immediately instruction
planted at the sub-object's vtable slot 2 - a bx lr gadget at 0x37510 - to neutralise the
virtual copy so Execute could reach the concatenation and the system() call without a fully
constructed engine. It returned 1 and wrote a 39-byte root-owned marker containing
uid=0(root) gid=0(root) groups=0(root). Two arithmetic checks corroborate rather than
merely assert: the content line plus trailing newline is exactly 39 bytes, and the
wide-conversion lengths independently confirm the payload loaded into the object - 1 for
the single-character command, whose first wide character is 0x78, and 41 for the parameter,
whose first three wide characters are 0x3b, 0x69 and 0x64, i.e. semicolon, i, d - proving
the injection prefix survived the narrow-to-wide conversion intact rather than being
encoded or escaped. A control run with a benign 25-character parameter reached the same
sink, returned the same value and wrote nothing, which is what upgrades this from "we called
a function and a file appeared" to "the file appeared because of the injected statement".
END-TO-END HTTP EXPLOITATION WAS NOT PERFORMED, NO REQUEST WAS EVER DELIVERED TO
pwrstudio, AND NO PHYSICAL DEVICE WAS USED. The unauthenticated write and the unauthenticated
restart are proven statically, from the route table, the handler decompilations and the
caller census; they are not
Severity
No CVSS data available.
Assigner
References
7 references
{
"containers": {
"cna": {
"affected": [
{
"product": "LineEds",
"vendor": "Circutor",
"versions": [
{
"status": "affected",
"version": "unknown"
}
]
}
],
"credits": [
{
"lang": "en",
"type": "finder",
"value": "disclosure via Fulldisclosure"
}
],
"descriptions": [
{
"lang": "en",
"value": "0day Rubbish Research Team is publicly disclosing a vulnerability in the pwrstudio\ndaemon shipped in Circutor LineEds (Line Energy Data System) industrial energy gateway\nfirmware 24.11.14-r0, engine variant pss0. Circutor S.A. is in Spain; the appliance sits\nbetween metering and power-quality instrumentation on one side and an energy-management\nor SCADA back office on the other.\n\nType: operating-system command injection (CWE-78) reached without authentication\n(CWE-306). Those two are the classes recorded in the research; the advisory\u0027s own\nclassification additionally maps CWE-285, because the permission gate fails open in two\nindependent ways, and CWE-20, because the events document is accepted with no schema\nvalidation, as contributing weaknesses, with CWE-250/CWE-269 recorded as bearing on\nremediation priority rather than on the defect.\n\nRoot cause: the daemon\u0027s event engine loads its configuration from events.xml, exposed at\n/services/user/events.xml over the management HTTP interface. An event may carry a\nshellExecute action whose command and parameter elements are concatenated by\nXCEventActionShellExecute::Execute (FUN_00548414) with the wide-string format \"%ls %ls\"\nstored at 0x7cc2d0, converted from wide to narrow, and passed to system() at call site\n0x5484bc. Nothing between the socket and the sink neutralises shell metacharacters: no\nquoting, no filtering, no allowlist. The format string is the whole reason the injection\nworks the way it does - two %ls conversions separated by a single space mean the sink\nperforms concatenation rather than argument separation, so command and parameter become one\nshell word sequence with no quoting boundary between them; a payload needs only a statement\nseparator, not a quote escape. The defect is at that concatenation rather than in the XML\nparser, which performs a pure element-to-field copy and is behaving correctly.\n\nExploitation is two HTTP requests with no delivery precondition of any kind. Step one\nwrites an event definition into /services/user/events.xml carrying an always-true\ncondition so it enters the live set on every evaluation pass, a deliberately invalid\ncommand token so the first shell statement fails harmlessly, and a parameter of the form\nsemicolon, command, semicolon, hash - yielding a shell line such as\n\"x ;id \u003e/tmp/circutor_eds_vuln_rce_marker;#\". Step two requests the restart.xml route,\nwhich makes the engine reload the document through XCEventDriver::LoadXml (FUN_0020acac),\nbuild the action object, evaluate the condition and dispatch Execute. There is no race, no\nmemory-layout dependency, no file to pre-place on the target and no dependence on another\nvulnerability, and because the intermediate state is a file the product itself persists,\nthe two steps can be separated in time arbitrarily. The binary is compiled with\nCONFIG_NO_HASP, so no licence-dongle gate stands between a written event and the shell\naction. The image is a non-PIE EXEC, so every offset quoted is a fixed file-and-memory\naddress reproducible against the same build with no relocation arithmetic.\n\nUnauthenticated classification, and it does not rest on a configuration assumption. The\nproduct implements authorization as a per-handler gate rather than as a dispatcher\nmiddleware stage: the gate is one function, FUN_001a93cc, invoked with a permission\nconstant by whichever handler chooses to invoke it. A cross-reference census of that\nfunction in the shipped binary records 197 call sites, all individual action handlers, all\ninside the address range 0x1aexxx to 0x1bfxxx. Both handlers in this chain fall outside it:\nthe restart handler FUN_001e3554 at 0x1e3554 and the events write handler FUN_0023ace0 at\n0x23ace0. The gate is therefore structurally incapable of authorizing either, because it is\nnever invoked on their paths. The restart handler\u0027s only guard is FUN_001a955c, a\nthree-instruction predicate testing that the request-context fields at +0x34 and +0x4c are\nboth non-null; those are the request and response object pointers installed\nunconditionally by the dispatcher\u0027s own setup routine FUN_001add40, which the route entry\nat 0x76a480 names at offset +0x0c, ahead of the handler at +0x10. It tests object presence,\nnot credentials. One honest tension is published rather than smoothed over: the events write\nhandler and its dispatcher contain no authorization call that we could find, yet the handler\nis recorded as capable of both 401 and 200 outcomes and we could not determine what selects\nthe 401 branch - so a credential-required conditional score is carried alongside the\nprimary rating precisely to account for that unresolved path.\n\nA second, configuration-independent defect is documented in the same advisory. The gate\u0027s\nown logic returns success with no check of any kind when the permission constant is not\n0x15, and returns success when enforcement is toggled off, so it fails open in two\nindependent ways for the 197 handlers that do call it. This chain does not depend on that\ndefect, and the advisory deliberately makes no claim that any specific other route is a\nsecond path to command execution; it is reported as a separate finding for vendor review.\n\nTwo claims carried by the research record are deliberately NOT published as fact. First,\nthat security enforcement is off by default: no shipped artifact supports the default value\n- nothing inside the image, no value read from the extracted root filesystem, not the\ncontents of the engine.xml package, and no vendor documentation of the factory setting. The\nunauthenticated classification does not depend on it, because neither route in the chain\ninvokes the gate at all; the default matters only for the alternative reloadCfg.xml\ntrigger, whose reload action does call the gate with permission 0x15, and that variant is\ntreated as conditional. Second, that pwrstudio runs as root on a device: the observed\nuid=0(root) is the privilege of the root-run emulation harness, and the device-side account\nis inferred from the daemon\u0027s ownership of the restart, reload and upgrade routes, since no\ninit script or service unit was extracted. Neither caveat alters the impact metrics.\n\nScoring. Four readings were priced and all four are published, two of them as not adopted:\n- PRIMARY, 9.8 Critical, CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H\n- CONDITIONAL, 8.8 High, CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H, applying where a\n deployment rejects anonymous configuration writes - enforcement enabled, or the events\n handler\u0027s unresolved 401 branch active - so that any valid operator credential suffices.\n Nothing else in the chain changes.\n- NOT ADOPTED, 10.0 Critical, CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H - the Scope\n Changed reading a reviewer who counts the appliance OS as a realm distinct from the\n management daemon would apply. We do not adopt it: the command executes inside the\n authority the vulnerable component already holds and no sandbox, container,\n virtualisation boundary or separate security realm is crossed. Published so the choice is\n visible and so the higher number is not mistaken for our claim.\n- NOT ADOPTED, 8.1 High, CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H - Attack Complexity\n raised despite the absence of any attacker-unsatisfiable precondition.\nA PR:H reading of 7.2 is not carried, because nothing indicates the configuration write or\nthe restart route demands an administrator rather than any authenticated principal.\n\nOn scoring accuracy: the research record carried 9.8 with that primary vector and\nrecomputing it under the CVSS 3.1 base algorithm yields 9.8, so the recorded number is\narithmetically correct. The reconciliation that matters is the grounding rather than the\narithmetic: the record derived PR:N from an asserted factory default for the security\ntoggle, for which no shipped artifact exists, whereas the advisory derives PR:N from the\nhandler call census and the two handler decompilations, which are artifact evidence. The\nscore is unchanged and the reason for it has been replaced with one that survives scrutiny.\nWe note this because several earlier advisories in this repository carried recorded scores\nthat did not compute for their recorded vectors.\n\nImpact: arbitrary command execution on an industrial energy gateway with the daemon\u0027s\nprivileges. Confidentiality High - every measurement record, configuration item, credential,\ncertificate and key on the appliance becomes readable, including whatever it holds to\nauthenticate to the energy-management back office, and a compromised gateway is an ideal\npassive collection point because it sees the traffic it forwards. Integrity High -\nmeasurement data can be altered in transit or at rest before reaching the back office, so\nforged consumption, generation or power-quality figures propagate into reporting,\nsettlement, alarm thresholds and control decisions while arriving through the legitimate\npath and are hard to distinguish from real data; event configuration can be rewritten and\nfirmware pushed through the same management plane, since the upgrade route is served by the\nsame daemon. Availability High - the attacker controls the very restart and reload\nprimitives this chain uses, and can reboot on demand, stop measurement collection, drop\nconfiguration or wedge the engine, producing a data outage in the metering path.\nPersistence and lateral movement are High in practice on the inference below: a foothold\ninside the OT segment on a device that security tooling normally treats as trusted\ninstrumentation.\n\nVerification boundary, stated plainly. Confirmation used the shipped binary extracted from\nthe firmware package, run under qemu-arm-static USER-MODE emulation with gdb - no\nfull-system emulation, so no kernel, device drivers, network stack or reverse-proxy front\nend were present. No physical device was used at any point: nothing was powered, connected,\nflashed or queried. The daemon\u0027s HTTP stack could not be started under emulation, for a\nspecific reason: its serial initialisation thread loops on an open of a Freescale i.MX\nprivate serial device returning EACCES, the device does not exist under qemu-user and no\ndevice model is provided, so the engine-ready futex is never signalled and the worker thread\nthat would accept connections is never scheduled. A listening socket on port 1500 does come\nup with an accept backlog, but nothing ever issues accept or poll against it, so no request\ncan be delivered and no response produced.\n\nThe real shipped sink function was therefore invoked directly under debugger control with\nthe field values the documented XML schema produces, using a return-immediately instruction\nplanted at the sub-object\u0027s vtable slot 2 - a bx lr gadget at 0x37510 - to neutralise the\nvirtual copy so Execute could reach the concatenation and the system() call without a fully\nconstructed engine. It returned 1 and wrote a 39-byte root-owned marker containing\nuid=0(root) gid=0(root) groups=0(root). Two arithmetic checks corroborate rather than\nmerely assert: the content line plus trailing newline is exactly 39 bytes, and the\nwide-conversion lengths independently confirm the payload loaded into the object - 1 for\nthe single-character command, whose first wide character is 0x78, and 41 for the parameter,\nwhose first three wide characters are 0x3b, 0x69 and 0x64, i.e. semicolon, i, d - proving\nthe injection prefix survived the narrow-to-wide conversion intact rather than being\nencoded or escaped. A control run with a benign 25-character parameter reached the same\nsink, returned the same value and wrote nothing, which is what upgrades this from \"we called\na function and a file appeared\" to \"the file appeared because of the injected statement\".\n\nEND-TO-END HTTP EXPLOITATION WAS NOT PERFORMED, NO REQUEST WAS EVER DELIVERED TO\npwrstudio, AND NO PHYSICAL DEVICE WAS USED. The unauthenticated write and the unauthenticated\nrestart are proven statically, from the route table, the handler decompilations and the\ncaller census; they are not"
}
],
"problemTypes": [
{
"descriptions": [
{
"cweId": "CWE-20",
"description": "CWE-20",
"lang": "en",
"type": "CWE"
},
{
"cweId": "CWE-250",
"description": "CWE-250",
"lang": "en",
"type": "CWE"
},
{
"cweId": "CWE-269",
"description": "CWE-269",
"lang": "en",
"type": "CWE"
},
{
"cweId": "CWE-285",
"description": "CWE-285",
"lang": "en",
"type": "CWE"
},
{
"cweId": "CWE-306",
"description": "CWE-306",
"lang": "en",
"type": "CWE"
},
{
"cweId": "CWE-78",
"description": "CWE-78",
"lang": "en",
"type": "CWE"
}
]
}
],
"providerMetadata": {
"dateUpdated": "2026-11-02T07:30:01Z",
"orgId": "4e2abfbf-4a2a-4b76-a4e0-d77c18ba156c",
"shortName": "VULNARCHIVE"
},
"references": [
{
"tags": [
"technical-description",
"exploit"
],
"url": "https://vuln.freearchive.org/archive/full-disclosure/2026/Oct/4"
},
{
"tags": [
"technical-description"
],
"url": "https://seclists.org/fulldisclosure/2026/Oct/4"
},
{
"url": "https://0day-rubbish.com"
},
{
"url": "https://0day-rubbish.com/blog/circutor-lineeds-pwrstudio-shellexecute-unauth-command-injection"
},
{
"url": "https://github.com/Exploit-Garbage/0day-Rubbish"
},
{
"url": "https://nmap.org/mailman/listinfo/fulldisclosure"
},
{
"url": "https://seclists.org/fulldisclosure/"
}
],
"source": {
"defect": [
"https://seclists.org/fulldisclosure/2026/Oct/4"
],
"discovery": "EXTERNAL"
},
"title": "[0day-rubbish] Circutor LineEds 24.11.14-r0 unauthenticated pwrstudio events.xml shellExecute command injection (9.8)",
"x_gcve": [
{
"recordType": "advisory",
"relationships": [],
"vulnId": "GCVE-1988-2026-0467",
"x_vulnarchive": {
"archiveUrl": "https://vuln.freearchive.org/archive/full-disclosure/2026/Oct/4",
"automated": true,
"contentSha256": "2296534ccbe441926f8e0393af24df4526cdcf42cb04005ba8eebd60c1329856",
"evidenceScore": 11,
"messageId": "",
"originalUrl": "https://seclists.org/fulldisclosure/2026/Oct/4",
"policy": "vulnarchive-1",
"sourceFormat": "text/html",
"sourcePublishedAt": "2026-10-05T17:52:58Z"
}
}
]
}
},
"cveMetadata": {
"assignerOrgId": "4e2abfbf-4a2a-4b76-a4e0-d77c18ba156c",
"assignerShortName": "VULNARCHIVE",
"datePublished": "2026-11-02T07:30:01Z",
"dateUpdated": "2026-11-02T07:30:01Z",
"state": "PUBLISHED",
"vulnId": "GCVE-1988-2026-0467"
},
"dataType": "CVE_RECORD",
"dataVersion": "5.2"
}