Monthly
Argument injection in the Claude Agent SDK for Python (claude-agent-sdk) before 0.2.121 allows remote attackers to achieve arbitrary command execution when an embedding application passes untrusted input into the SDK's `resume` option. Because `--resume` is declared by the Claude CLI with an optional value, a session identifier beginning with `-` is parsed as a separate CLI flag rather than the option's value, letting an attacker inject `--mcp-config` and register a stdio MCP server that executes arbitrary commands at CLI startup while bypassing the SDK's permission controls. The assessed CVSS:3.1 vector (AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H) indicates no authentication is required but exploitation is high-complexity and conditional on the host application forwarding attacker-controlled data as the session ID; a vendor patch is available as version 0.2.121, and no public exploit code was identified at time of analysis.
Improper input validation in Google Chrome Desktop versions prior to 155.0.8059.84 could let a remote attacker execute arbitrary code outside the browser sandbox when a victim loads a crafted HTML page. The CVSS 3.1 vector scores 9.6 (AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H), meaning no authentication is required but the victim must interact with attacker-supplied content, and the Changed Scope rating implies a security-boundary crossing such as an escape from the renderer sandbox - a non-trivial step that CISA's SSVC rates as not automatable. Despite the critical base score, the practical risk signal is mixed at time of analysis: exploitation status is none per SSVC, no public exploit code has been identified, and EPSS is only 0.21% (11th percentile), while Chromium's own severity classification for this issue is Medium and a vendor patch is available.
Paid fulfillment bypass in Lychee (Supporter/Pro editions) before 7.7.0: the checkout validators accept any OmnipayProviderType enum value, including the Dummy test gateway that is deliberately hidden from production checkout, so a remote unauthenticated buyer - where webshop_allow_guest_checkout is enabled - can start and finalize a session against the Dummy gateway and, with webshop_auto_fulfill_enabled on, have OrderCompletedListener::fullfillOrder() release the purchasable item without any real payment. The integrity impact is genuine but conditional (CVSS 4.0 base 8.3 with Attack Requirements: Present; assessed CVSS 3.1 8.3, C:L/I:H/A:N), since exploitation requires a paid Supporter/Pro instance running the webshop with at least one enabled purchasable item and the relevant opt-in flags. No public exploit identified at time of analysis, and the vendor has released a fixed version, 7.7.0.
Unauthenticated buyers on Lychee instances running the Supporter/Pro webshop can tamper with the parameters handed to the payment gateway, overriding the server-computed amount, currency, returnUrl, cancelUrl, or transactionId so that a gateway approves a payment for an attacker-chosen price while the merchant is underpaid and auto-fulfilled goods are still delivered. The flaw lives in CheckoutService::preparePurchaseParameters(), which merges the client-supplied additional_data array after the server's own purchase parameters, allowing later string keys to overwrite earlier ones before processPayment() reaches a non-PayPal Omnipay gateway. Exploitation is not a network-wide exposure: it requires an enabled webshop, at least one purchasable item, and an affected Omnipay gateway (PayPal is explicitly not affected), which sharply narrows the population at risk; no public exploit code was identified at time of analysis, though the upstream fix ships a regression test that demonstrates the parameter override, and no CISA KEV entry exists. This is best treated as a credible, prioritizable business-logic flaw rather than an inflated high-severity issue, since the integrity impact on the merchant's billing is real but the prerequisites are opt-in and non-default.
Local, authenticated attackers with low privileges can cause process termination or memory corruption in HPE NonStop SQL/MX UDR Server by sending a crafted message. The vulnerability is an improper input validation (CWE-20) issue in the UDR SERVER component that requires local access and a running UDR process. No public exploit identified at time of analysis; the primary impact is denial of service (availability high) with a low integrity impact and no confidentiality impact.
Unauthorized command execution in HPE NonStop SQL/MX SQLCAT allows remote attackers to run arbitrary commands through improperly validated input (CWE-20). The CVSS score is 7.4 with a vector indicating low-privileged authenticated access (PR:L), yet the HPE description states an unauthenticated user can exploit it, creating an unresolved contradiction that complicates prioritization. No public exploit code or active exploitation has been identified at time of analysis, and the risk is confined to HPE NonStop environments where the SQLCAT catalog manager is network-accessible.
Memory corruption in the HPE NonStop HTTP WEBSERVER can be triggered by network-delivered HTTP input, causing the web server process to fail and potentially halting the NonStop system itself (CVSS 8.2, AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:H). Any HPE NonStop deployment that exposes the WEBSERVER component to an untrusted network is in scope, since the assessed vector requires no authentication, privileges, or user interaction against a listening server. The practical risk is bounded by the relatively small installed base of the NonStop platform and by the fact that the worst realistic outcome is availability loss rather than code execution or data disclosure; there is no confirmed active exploitation (not in CISA KEV) and no public exploit code identified at time of analysis.
Memory corruption in HPE NonStop SOAP 4.0 can be triggered remotely by a crafted SOAP request, resulting in a crash that takes down the affected NonStop system or SOAP service. The vulnerability is network-reachable and requires no authentication or user interaction (CVSS:3.1/AV:N/AC:L/PR:N/UI:N), scoring 8.2 with the impact concentrated on availability (A:H) alongside a limited integrity effect (I:L) and no confidentiality exposure. No public exploit code or confirmed active exploitation has been identified at time of analysis, and the realistic risk depends heavily on whether the SOAP listener is exposed beyond a restricted internal back-office network, making this a moderate-priority availability issue rather than a top-tier emergency.
Out-of-bounds write in HPE NonStop QIOMON (T8671L02^ABH and earlier) allows a low-privilege authenticated actor to corrupt adjacent buffer memory when QIOMON processes attacker-influenced input, with the integrity and availability impact rated High and confidentiality impact Low. Exploitation is constrained on two fronts: the attacker must already hold authenticated low-privilege access to the NonStop environment (CVSS PR:L) and a legitimate user must perform an action that feeds malicious input into QIOMON (CVSS UI:R), so the issue is not exploitable as an end-to-end unauthenticated remote attack. No public exploit code has been identified at time of analysis and the CVE is not listed in CISA KEV; EPSS data was not provided, so it should be treated as a moderate-priority, non-emergency hardening item.
An out-of-bounds write in HPE NonStop CIPSAM (version T0693L02^BAU and earlier) lets an attacker with local access and at least low-level privileges corrupt adjacent buffer memory, producing High integrity and High availability impact on the affected NonStop system. The CVSS 3.1 base score of 8.7 is driven by the changed-scope metric combined with those impact values rather than by remote reach - the vector is AV:L/AC:L/PR:L/UI:N/S:C/C:L/I:H/A:H, so this is an authenticated, local-access issue and is not remotely exploitable by an unauthenticated attacker. No public exploit code and no confirmed active exploitation were identified at time of analysis, and the realistic attacker population is narrowed by the requirement for prior credentials on a proprietary, access-controlled mission-critical platform.
Argument injection in the Claude Agent SDK for Python (claude-agent-sdk) before 0.2.121 allows remote attackers to achieve arbitrary command execution when an embedding application passes untrusted input into the SDK's `resume` option. Because `--resume` is declared by the Claude CLI with an optional value, a session identifier beginning with `-` is parsed as a separate CLI flag rather than the option's value, letting an attacker inject `--mcp-config` and register a stdio MCP server that executes arbitrary commands at CLI startup while bypassing the SDK's permission controls. The assessed CVSS:3.1 vector (AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H) indicates no authentication is required but exploitation is high-complexity and conditional on the host application forwarding attacker-controlled data as the session ID; a vendor patch is available as version 0.2.121, and no public exploit code was identified at time of analysis.
Improper input validation in Google Chrome Desktop versions prior to 155.0.8059.84 could let a remote attacker execute arbitrary code outside the browser sandbox when a victim loads a crafted HTML page. The CVSS 3.1 vector scores 9.6 (AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H), meaning no authentication is required but the victim must interact with attacker-supplied content, and the Changed Scope rating implies a security-boundary crossing such as an escape from the renderer sandbox - a non-trivial step that CISA's SSVC rates as not automatable. Despite the critical base score, the practical risk signal is mixed at time of analysis: exploitation status is none per SSVC, no public exploit code has been identified, and EPSS is only 0.21% (11th percentile), while Chromium's own severity classification for this issue is Medium and a vendor patch is available.
Paid fulfillment bypass in Lychee (Supporter/Pro editions) before 7.7.0: the checkout validators accept any OmnipayProviderType enum value, including the Dummy test gateway that is deliberately hidden from production checkout, so a remote unauthenticated buyer - where webshop_allow_guest_checkout is enabled - can start and finalize a session against the Dummy gateway and, with webshop_auto_fulfill_enabled on, have OrderCompletedListener::fullfillOrder() release the purchasable item without any real payment. The integrity impact is genuine but conditional (CVSS 4.0 base 8.3 with Attack Requirements: Present; assessed CVSS 3.1 8.3, C:L/I:H/A:N), since exploitation requires a paid Supporter/Pro instance running the webshop with at least one enabled purchasable item and the relevant opt-in flags. No public exploit identified at time of analysis, and the vendor has released a fixed version, 7.7.0.
Unauthenticated buyers on Lychee instances running the Supporter/Pro webshop can tamper with the parameters handed to the payment gateway, overriding the server-computed amount, currency, returnUrl, cancelUrl, or transactionId so that a gateway approves a payment for an attacker-chosen price while the merchant is underpaid and auto-fulfilled goods are still delivered. The flaw lives in CheckoutService::preparePurchaseParameters(), which merges the client-supplied additional_data array after the server's own purchase parameters, allowing later string keys to overwrite earlier ones before processPayment() reaches a non-PayPal Omnipay gateway. Exploitation is not a network-wide exposure: it requires an enabled webshop, at least one purchasable item, and an affected Omnipay gateway (PayPal is explicitly not affected), which sharply narrows the population at risk; no public exploit code was identified at time of analysis, though the upstream fix ships a regression test that demonstrates the parameter override, and no CISA KEV entry exists. This is best treated as a credible, prioritizable business-logic flaw rather than an inflated high-severity issue, since the integrity impact on the merchant's billing is real but the prerequisites are opt-in and non-default.
Local, authenticated attackers with low privileges can cause process termination or memory corruption in HPE NonStop SQL/MX UDR Server by sending a crafted message. The vulnerability is an improper input validation (CWE-20) issue in the UDR SERVER component that requires local access and a running UDR process. No public exploit identified at time of analysis; the primary impact is denial of service (availability high) with a low integrity impact and no confidentiality impact.
Unauthorized command execution in HPE NonStop SQL/MX SQLCAT allows remote attackers to run arbitrary commands through improperly validated input (CWE-20). The CVSS score is 7.4 with a vector indicating low-privileged authenticated access (PR:L), yet the HPE description states an unauthenticated user can exploit it, creating an unresolved contradiction that complicates prioritization. No public exploit code or active exploitation has been identified at time of analysis, and the risk is confined to HPE NonStop environments where the SQLCAT catalog manager is network-accessible.
Memory corruption in the HPE NonStop HTTP WEBSERVER can be triggered by network-delivered HTTP input, causing the web server process to fail and potentially halting the NonStop system itself (CVSS 8.2, AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:H). Any HPE NonStop deployment that exposes the WEBSERVER component to an untrusted network is in scope, since the assessed vector requires no authentication, privileges, or user interaction against a listening server. The practical risk is bounded by the relatively small installed base of the NonStop platform and by the fact that the worst realistic outcome is availability loss rather than code execution or data disclosure; there is no confirmed active exploitation (not in CISA KEV) and no public exploit code identified at time of analysis.
Memory corruption in HPE NonStop SOAP 4.0 can be triggered remotely by a crafted SOAP request, resulting in a crash that takes down the affected NonStop system or SOAP service. The vulnerability is network-reachable and requires no authentication or user interaction (CVSS:3.1/AV:N/AC:L/PR:N/UI:N), scoring 8.2 with the impact concentrated on availability (A:H) alongside a limited integrity effect (I:L) and no confidentiality exposure. No public exploit code or confirmed active exploitation has been identified at time of analysis, and the realistic risk depends heavily on whether the SOAP listener is exposed beyond a restricted internal back-office network, making this a moderate-priority availability issue rather than a top-tier emergency.
Out-of-bounds write in HPE NonStop QIOMON (T8671L02^ABH and earlier) allows a low-privilege authenticated actor to corrupt adjacent buffer memory when QIOMON processes attacker-influenced input, with the integrity and availability impact rated High and confidentiality impact Low. Exploitation is constrained on two fronts: the attacker must already hold authenticated low-privilege access to the NonStop environment (CVSS PR:L) and a legitimate user must perform an action that feeds malicious input into QIOMON (CVSS UI:R), so the issue is not exploitable as an end-to-end unauthenticated remote attack. No public exploit code has been identified at time of analysis and the CVE is not listed in CISA KEV; EPSS data was not provided, so it should be treated as a moderate-priority, non-emergency hardening item.
An out-of-bounds write in HPE NonStop CIPSAM (version T0693L02^BAU and earlier) lets an attacker with local access and at least low-level privileges corrupt adjacent buffer memory, producing High integrity and High availability impact on the affected NonStop system. The CVSS 3.1 base score of 8.7 is driven by the changed-scope metric combined with those impact values rather than by remote reach - the vector is AV:L/AC:L/PR:L/UI:N/S:C/C:L/I:H/A:H, so this is an authenticated, local-access issue and is not remotely exploitable by an unauthenticated attacker. No public exploit code and no confirmed active exploitation were identified at time of analysis, and the realistic attacker population is narrowed by the requirement for prior credentials on a proprietary, access-controlled mission-critical platform.