Monthly
Heap out-of-bounds write in llama.cpp builds b5899 through b10260 lets an attacker corrupt heap memory when the victim loads a crafted GGUF model that uses the PLaMo2 tokenizer. A BYTE-typed token whose text parses to a negative value (for example <0x-1>) is used as an index into the tokenizer's bytes_ buffer during model loading, so the write lands before the buffer and can crash the process or potentially enable code execution. The flaw is local in nature (it triggers only when an untrusted model is loaded, per the assessed CVSS:3.1/AV:L/AC:L/PR:N/UI:R vector), a vendor patch is available in build b10261, no public exploit has been identified at time of analysis, and EPSS is low at 0.14% (3rd percentile) with a this_week priority label.
Out-of-vocabulary stop_token_ids in vLLM 0.20.1 through 0.28.x allow a client able to submit generation requests to terminate the inference engine, causing denial of service for every other user sharing that server; on the V2 runner the same unchecked index can instead write outside the logits buffer and corrupt adjacent logits or other GPU state. The flaw is an input-validation gap: SamplingParams accepted any integer for stop_token_ids and those values became column indices into the model's logits tensor, so the OpenAI chat and completion paths, the legacy generate endpoint, and batch inputs are all affected. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N/VA:H) puts this at 7.1 with availability-only impact, and the PR:L metric is the main limiting factor, meaning the attacker needs the privileged capability to submit requests rather than being fully anonymous; no public exploit code or active exploitation has been identified at time of analysis, and priority is this-week remediation rather than emergency response.
Velneo vServer can be knocked into a denial-of-service state by unauthenticated network traffic that exploits improper validation of an array index (CWE-129) in the server's request handling. Only installations running the 36.x branch before 36.6 or the 37.x branch before 37.2 are affected, and the assessed impact is availability-only (CVSS 4.0 8.7, VA:H with VC:N/VI:N) - there is no demonstrated path to data disclosure, tampering, or code execution from this flaw alone. Vendor fixes are published in Velneo's security advisories, and no public exploit code or active exploitation (CISA KEV) has been identified at time of analysis.
Remote MMS Read requests can crash libIEC61850 MMS servers running versions prior to 1.6.2 when the server's data model contains array-typed variables. An unauthenticated attacker (CVSS vector PR:N, AV:N) sends an alternate-array-access read with numberOfElements greater than zero and a sub-array range that exceeds the array bounds; the resulting out-of-range MmsValue_getElement call returns NULL, which is then passed to MmsValue_isDeletable and dereferenced, producing a NULL pointer dereference and process crash. Impact is availability-only (CVSS 7.5, C:N/I:N/A:H) and scoped to the array sub-access code path, since the single-element error path handles the same invalid index safely; no public exploit identified at time of analysis.
Out-of-bounds array access in the Svc::GenericHub component of NASA's F´ (F Prime) framework (4.1.0 through 4.1.x, fixed in 4.2.0) lets remote, unauthenticated attackers who can deliver a crafted hub packet via fromBufferDriver supply an output-port index outside the configured port array; assert-enabled builds abort (denial of service) while builds compiled without assertions can corrupt heap memory and potentially achieve remote code execution. No public exploit code exists and no exploitation has been confirmed in the wild - exploitation presupposes attacker reachability to the hub's input channel, which in F´ deployments is typically a mission radio link or inter-process transport rather than an internet-facing service. The vendor rates this CWE-129 improper array index validation flaw CVSS 4.0 9.3 (Critical), with high confidentiality, integrity and availability impact.
Heap buffer overflow in FreeRDP 3.0.0 through 3.31.x allows an attacker who controls a physical smartcard token to corrupt heap memory during PKCS#11 smartcard logon, potentially achieving code execution with the privileges of the FreeRDP client. The flaw is in collect_keys(), which enumerates public key objects across PKCS#11 slots into a fixed 64-entry array without bounds checking; a token exposing more than 64 keys with attacker-controlled CKA_ID and CKA_LABEL values overflows the array. Exploitation requires PKCS#11 smartcard authentication to be configured, physical presentation of the malicious token, and user-initiated logon, and no public exploit code or active exploitation is known at time of analysis; this is a constrained, physically proximate issue rather than a mass-exploitation priority.
Denial of service in the MongoDB Go Driver can be triggered when an application calls bson.RawArray.Validate or bsoncore.Array.Validate on a malformed four-byte BSON array supplied by an unauthenticated remote source, causing an out-of-bounds index and a runtime panic that may terminate the unprotected process. The CVE carries a CVSS 4.0 base score of 8.2, but our assessed CVSS 3.1 vector (AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H) reflects the prerequisite that the vulnerable Validate methods must be invoked on attacker-controlled raw BSON; applications that never call these methods on untrusted input or that recover from panics are not materially impacted. No public exploit identified at time of analysis, no confidentiality or integrity impact has been identified, and the realistic blast radius is availability of the specific unprotected process.
Go slice-bounds panic in the Excelize spreadsheet library versions 2.8.1 through 2.11.0 lets a crafted .xlsx workbook crash any process or request worker that calls GetPivotTables. The library parses pivot-table field indices separately from pivot-cache field names and indexes one slice with the other without bounds checks, so a workbook with a pivot-field count mismatch or an out-of-range/negative dataField fld value triggers an unrecovered runtime panic that escapes the library (CVSS 3.1 assessed 8.7, availability-only; no authentication or user interaction required from the caller's perspective). No fixed release version is confirmed in the advisory data - only the upstream fix commit - and no public exploit code was identified at time of analysis, though exploitation requires no more than delivering a malicious workbook to a service that parses pivot tables.
Local privilege escalation in the NVIDIA GPU Display Driver for Windows and Linux lets a low-privileged local user trigger an out-of-bounds array access in the kernel mode layer, potentially yielding code execution in kernel context, denial of service, escalation of privileges, information disclosure and data tampering. The assessed CVSS 3.1 vector is AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8), meaning the attack is strictly local and requires only low-privileged authenticated code on the host, with no user interaction and low complexity - so any unprivileged process able to reach the affected kernel-mode interface on a system running GeForce, RTX/Quadro/NVS or Tesla drivers (or a vGPU host reachable from a guest) can attempt it. There is confirmed actively exploited (CISA KEV) status: no - the CVE is not in KEV and no public exploit code has been identified at time of analysis, so early patching is key while exploitation activity has not yet been observed.
Denial of service in vLLM before 0.29.0 allows an authenticated API user to permanently disrupt the inference service with a single malformed request to the OpenAI-compatible POST /v1/completions or POST /v1/chat/completions endpoints. The server accepts user-supplied stop_token_ids and checks only that they are integers, not that they fall within the model's vocabulary/logits range; when min_tokens > 0 those ids are used as logits indices to suppress stop tokens, so an out-of-range value reaches a CUDA index_put_ operation and trips a device-side assertion that returns HTTP 500 and leaves EngineCore in a fatal state until the service is restarted. The impact is bounded to availability (no code execution or data disclosure) and exploitation requires valid API credentials plus exposure of the OpenAI-compatible endpoints; no public exploit code was identified at time of analysis, and the vendor has released a fix.
Heap out-of-bounds write in llama.cpp builds b5899 through b10260 lets an attacker corrupt heap memory when the victim loads a crafted GGUF model that uses the PLaMo2 tokenizer. A BYTE-typed token whose text parses to a negative value (for example <0x-1>) is used as an index into the tokenizer's bytes_ buffer during model loading, so the write lands before the buffer and can crash the process or potentially enable code execution. The flaw is local in nature (it triggers only when an untrusted model is loaded, per the assessed CVSS:3.1/AV:L/AC:L/PR:N/UI:R vector), a vendor patch is available in build b10261, no public exploit has been identified at time of analysis, and EPSS is low at 0.14% (3rd percentile) with a this_week priority label.
Out-of-vocabulary stop_token_ids in vLLM 0.20.1 through 0.28.x allow a client able to submit generation requests to terminate the inference engine, causing denial of service for every other user sharing that server; on the V2 runner the same unchecked index can instead write outside the logits buffer and corrupt adjacent logits or other GPU state. The flaw is an input-validation gap: SamplingParams accepted any integer for stop_token_ids and those values became column indices into the model's logits tensor, so the OpenAI chat and completion paths, the legacy generate endpoint, and batch inputs are all affected. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N/VA:H) puts this at 7.1 with availability-only impact, and the PR:L metric is the main limiting factor, meaning the attacker needs the privileged capability to submit requests rather than being fully anonymous; no public exploit code or active exploitation has been identified at time of analysis, and priority is this-week remediation rather than emergency response.
Velneo vServer can be knocked into a denial-of-service state by unauthenticated network traffic that exploits improper validation of an array index (CWE-129) in the server's request handling. Only installations running the 36.x branch before 36.6 or the 37.x branch before 37.2 are affected, and the assessed impact is availability-only (CVSS 4.0 8.7, VA:H with VC:N/VI:N) - there is no demonstrated path to data disclosure, tampering, or code execution from this flaw alone. Vendor fixes are published in Velneo's security advisories, and no public exploit code or active exploitation (CISA KEV) has been identified at time of analysis.
Remote MMS Read requests can crash libIEC61850 MMS servers running versions prior to 1.6.2 when the server's data model contains array-typed variables. An unauthenticated attacker (CVSS vector PR:N, AV:N) sends an alternate-array-access read with numberOfElements greater than zero and a sub-array range that exceeds the array bounds; the resulting out-of-range MmsValue_getElement call returns NULL, which is then passed to MmsValue_isDeletable and dereferenced, producing a NULL pointer dereference and process crash. Impact is availability-only (CVSS 7.5, C:N/I:N/A:H) and scoped to the array sub-access code path, since the single-element error path handles the same invalid index safely; no public exploit identified at time of analysis.
Out-of-bounds array access in the Svc::GenericHub component of NASA's F´ (F Prime) framework (4.1.0 through 4.1.x, fixed in 4.2.0) lets remote, unauthenticated attackers who can deliver a crafted hub packet via fromBufferDriver supply an output-port index outside the configured port array; assert-enabled builds abort (denial of service) while builds compiled without assertions can corrupt heap memory and potentially achieve remote code execution. No public exploit code exists and no exploitation has been confirmed in the wild - exploitation presupposes attacker reachability to the hub's input channel, which in F´ deployments is typically a mission radio link or inter-process transport rather than an internet-facing service. The vendor rates this CWE-129 improper array index validation flaw CVSS 4.0 9.3 (Critical), with high confidentiality, integrity and availability impact.
Heap buffer overflow in FreeRDP 3.0.0 through 3.31.x allows an attacker who controls a physical smartcard token to corrupt heap memory during PKCS#11 smartcard logon, potentially achieving code execution with the privileges of the FreeRDP client. The flaw is in collect_keys(), which enumerates public key objects across PKCS#11 slots into a fixed 64-entry array without bounds checking; a token exposing more than 64 keys with attacker-controlled CKA_ID and CKA_LABEL values overflows the array. Exploitation requires PKCS#11 smartcard authentication to be configured, physical presentation of the malicious token, and user-initiated logon, and no public exploit code or active exploitation is known at time of analysis; this is a constrained, physically proximate issue rather than a mass-exploitation priority.
Denial of service in the MongoDB Go Driver can be triggered when an application calls bson.RawArray.Validate or bsoncore.Array.Validate on a malformed four-byte BSON array supplied by an unauthenticated remote source, causing an out-of-bounds index and a runtime panic that may terminate the unprotected process. The CVE carries a CVSS 4.0 base score of 8.2, but our assessed CVSS 3.1 vector (AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H) reflects the prerequisite that the vulnerable Validate methods must be invoked on attacker-controlled raw BSON; applications that never call these methods on untrusted input or that recover from panics are not materially impacted. No public exploit identified at time of analysis, no confidentiality or integrity impact has been identified, and the realistic blast radius is availability of the specific unprotected process.
Go slice-bounds panic in the Excelize spreadsheet library versions 2.8.1 through 2.11.0 lets a crafted .xlsx workbook crash any process or request worker that calls GetPivotTables. The library parses pivot-table field indices separately from pivot-cache field names and indexes one slice with the other without bounds checks, so a workbook with a pivot-field count mismatch or an out-of-range/negative dataField fld value triggers an unrecovered runtime panic that escapes the library (CVSS 3.1 assessed 8.7, availability-only; no authentication or user interaction required from the caller's perspective). No fixed release version is confirmed in the advisory data - only the upstream fix commit - and no public exploit code was identified at time of analysis, though exploitation requires no more than delivering a malicious workbook to a service that parses pivot tables.
Local privilege escalation in the NVIDIA GPU Display Driver for Windows and Linux lets a low-privileged local user trigger an out-of-bounds array access in the kernel mode layer, potentially yielding code execution in kernel context, denial of service, escalation of privileges, information disclosure and data tampering. The assessed CVSS 3.1 vector is AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H (7.8), meaning the attack is strictly local and requires only low-privileged authenticated code on the host, with no user interaction and low complexity - so any unprivileged process able to reach the affected kernel-mode interface on a system running GeForce, RTX/Quadro/NVS or Tesla drivers (or a vGPU host reachable from a guest) can attempt it. There is confirmed actively exploited (CISA KEV) status: no - the CVE is not in KEV and no public exploit code has been identified at time of analysis, so early patching is key while exploitation activity has not yet been observed.
Denial of service in vLLM before 0.29.0 allows an authenticated API user to permanently disrupt the inference service with a single malformed request to the OpenAI-compatible POST /v1/completions or POST /v1/chat/completions endpoints. The server accepts user-supplied stop_token_ids and checks only that they are integers, not that they fall within the model's vocabulary/logits range; when min_tokens > 0 those ids are used as logits indices to suppress stop tokens, so an out-of-range value reaches a CUDA index_put_ operation and trips a device-side assertion that returns HTTP 500 and leaves EngineCore in a fatal state until the service is restarted. The impact is bounded to availability (no code execution or data disclosure) and exploitation requires valid API credentials plus exposure of the OpenAI-compatible endpoints; no public exploit code was identified at time of analysis, and the vendor has released a fix.