vLLM versions 0.5.5 through 0.17.x use incorrect mono audio downmixing via numpy.mean instead of the ITU-R BS.775-4 weighted standard, causing audio processed by AI models to diverge from human perception. An authenticated remote attacker with low privileges can exploit this inconsistency to manipulate audio-based model outputs or infer mismatches between expected and actual audio processing, affecting integrity of audio-driven inference pipelines. The vulnerability has been patched in vLLM 0.18.0.
Rack::Sendfile in versions prior to 2.2.23, 3.1.21, and 3.2.6 allows remote attackers to inject regex metacharacters into X-Accel-Mapping request headers, enabling unescaped interpolation that manipulates the X-Accel-Redirect response header and causes nginx to serve unintended files from internal locations. No public exploit code or active exploitation has been confirmed; patch versions are available from the vendor.
Leancrypto library prior to version 1.7.1 allows remote attackers to impersonate X.509 certificate identities by crafting certificates with padded Common Names that exploit integer overflow when casting size_t to uint8_t, enabling spoofing in PKCS#7 verification, certificate chain matching, and code signing scenarios. The vulnerability has a moderate CVSS score of 5.9 (network-accessible, high complexity attack) and is not confirmed in active exploitation, though the attack is technically straightforward once a malicious certificate is crafted.
Server-side request forgery (SSRF) in priyankark a11y-mcp up to version 1.0.5 allows local authenticated attackers to perform arbitrary outbound requests via the A11yServer function in src/index.js, potentially enabling access to internal services or exfiltration of sensitive data. The vulnerability requires local access and user approval (as the tool operates as a local stdio MCP server with no network exposure), and publicly available exploit code exists. Vendor has released patched version 1.0.6 with commit e3e11c9e8482bd06b82fd9fced67be4856f0dffc.
CocoaMQTT library versions prior to 2.2.2 allow remote denial of service when parsing malformed MQTT packets from a broker, causing immediate application crashes on iOS, macOS, and tvOS devices. An attacker or compromised MQTT broker can publish a 4-byte malformed payload with the RETAIN flag to persist it indefinitely, ensuring every vulnerable client that subscribes receives the crash-inducing packet, effectively bricking the application until manual intervention on the broker. The vulnerability requires an authenticated user context (PR:L in CVSS vector) but impacts application availability with high severity; patch version 2.2.2 is available.
Heap-based buffer overflow in Cesanta Mongoose versions up to 7.20 allows unauthenticated remote attackers to compromise confidentiality, integrity, and availability through malicious TLS 1.3 handshake manipulation. The vulnerability resides in mg_tls_recv_cert() function's improper handling of the pubkey argument during certificate processing. Publicly available exploit code exists (CVSS temporal E:P), and vendor-released patch is available in version 7.21. CVSS base score 7.3 reflects network-accessible, low-complexity attack requiring no privileges or user interaction.
Server-side request forgery in huimeicloud hm_editor up to version 2.2.3 allows remote attackers to manipulate the url parameter in the image-to-base64 endpoint (client.get function in src/mcp-server.js), enabling arbitrary HTTP requests from the server. The vulnerability carries a CVSS 6.9 score and publicly available exploit code exists; the vendor has not responded to early disclosure attempts.
SQL injection in itsourcecode Online Enrollment System 1.0 allows remote unauthenticated attackers to execute arbitrary SQL queries via the deptid parameter in /enrollment/index.php?view=edit&id=3, potentially enabling unauthorized data access, modification, or deletion. Publicly available exploit code exists, increasing real-world exploitation risk despite the moderate CVSS score of 6.9. The vulnerability affects the Parameter Handler component's SQL query construction logic.
Unblinded BPF immediate values in PROBE_MEM32 stores bypass constant hardening in the Linux kernel BPF JIT compiler when bpf_jit_harden >= 1, allowing user-controlled 32-bit immediates to leak into native code. The vulnerability affects Linux kernel versions where convert_ctx_accesses() rewrites arena pointer stores to BPF_ST|BPF_PROBE_MEM32 before constant blinding runs, but bpf_jit_blind_insn() only handles BPF_ST|BPF_MEM instructions. No public exploit code or active exploitation has been ide
Linux kernel TLS subsystem leaks socket buffers (skbs) when asynchronous AEAD decryption operations fail during batch processing, allowing local attackers to exhaust kernel memory and potentially cause denial of service. The vulnerability exists in tls_decrypt_async_wait() and related functions that manage the async_hold queue, which pins encrypted input buffers for AEAD engine references; improper cleanup in failure paths leaves these buffers allocated indefinitely. This is a kernel memory leak affecting TLS decryption in the kernel's cryptographic stack, confirmed by multiple upstream patches across stable branches.
Memory sealing (mseal) in the Linux kernel incorrectly tracks virtual memory area (VMA) boundaries during merge operations, causing curr_end to become stale and resulting in incorrect iteration state. This flaw in mm/mseal.c affects Linux kernel versions where the mseal feature is present, allowing local attackers to potentially bypass memory sealing protections or trigger information disclosure by manipulating VMA merge behavior during seal operations.
Stored cross-site scripting (XSS) in Hoppscotch prior to version 2026.3.0 allows authenticated users to inject malicious scripts via the team member display name field, which executes when other users view the overflow tooltip. The vulnerability requires user interaction (viewing the tooltip) and affects the confidentiality and integrity of affected sessions with a CVSS score of 5.4. No public exploit code or active exploitation has been identified at time of analysis.
Listmonk versions 4.1.0 through 6.0.x contain authorization bypass vulnerabilities in list permission checks that allow authenticated users in multi-user environments to access mailing lists they should not have access to. The vulnerability affects only self-hosted deployments with multiple untrusted users and has been patched in version 6.1.0. No public exploit code or active exploitation has been identified at this time.
Server-side request forgery (SSRF) in Postiz prior to version 2.21.4 allows authenticated users to create webhooks pointing to internal or private network addresses, which are then fetched without runtime validation when posts are published, enabling blind SSRF attacks against internal services. The vulnerability stems from inconsistent input validation: the webhook creation endpoint (POST /webhooks/) uses only basic URL format checking, while the update and test endpoints correctly enforce strict URL validation. CVSS 5.4 with EPSS exploitation probability reflects the requirement for authentication and limited direct impact, though the ability to target internal infrastructure represents meaningful risk.
Authenticated stored cross-site scripting (XSS) in IBM Content Navigator versions 3.0.15, 3.1.0, and 3.2.0 allows logged-in users to inject arbitrary JavaScript into the Web UI, potentially enabling credential theft or session hijacking within trusted browser contexts. The vulnerability requires user interaction (UI:R) and authenticated access (PR:L), limiting exposure to insider threats and social engineering scenarios where victims click attacker-controlled links. No public exploit code or active exploitation has been confirmed; patch availability from IBM mitigates the risk for patched deployments.
SEPPmail Secure Email Gateway before version 15.0.3 allows remote attackers to conceal security tags from end users by submitting emails with excessively long subject lines, enabling potential phishing and message spoofing attacks by bypassing visual security indicators. The vulnerability affects all versions prior to 15.0.3 across the product line and has been reported by Switzerland's National Cyber Security Centre (NCSC.ch). No active exploitation has been confirmed, and no public exploit code is currently available.
SEPPmail Secure Email Gateway before version 15.0.3 permits attackers to craft malicious password-tags that circumvent subject line sanitization controls, potentially enabling unauthorized email manipulation or information disclosure. The vulnerability affects all versions prior to 15.0.3 and was reported by NCSC.ch; no CVSS score or public exploit code has been published at the time of analysis.
SEPPmail Secure Email Gateway before version 15.0.3 allows attackers to upload PGP keys with mismatched User IDs and email addresses, enabling spoofing and potential information disclosure by circumventing email authentication controls. The vulnerability affects all versions prior to 15.0.3 and was reported by NCSC.ch. No public exploit code or active exploitation has been confirmed at the time of analysis.
SEPPmail Secure Email Gateway before version 15.0.3 allows unauthenticated external users to modify GINA webdomain metadata and bypass per-domain restrictions, enabling attackers to circumvent email security controls across isolated security domains. NCSC.ch reported this authentication bypass vulnerability affecting all versions prior to 15.0.3. The ability to modify domain-level metadata across organizational security boundaries represents a direct compromise of the gateway's core function to enforce per-domain policies.
Open edX Platform from maple release through ulmo allows unauthenticated attackers to bypass email verification by exploiting an OAuth2 password grant that issues tokens to inactive users combined with exposure of activation keys in the REST API response at /api/user/v1/accounts/. This authentication bypass enables account takeover and unauthorized access to learning platforms. The vulnerability affects all deployments from maple to before ulmo release and has been patched in the ulmo release.
OpenClaw versions prior to commit 8aceaf5 allow authenticated remote attackers to bypass shell-bleed protection validation by crafting complex command forms such as piped execution, command substitution, or subshell invocation, enabling execution of arbitrary script content that should be blocked. The vulnerability affects the validateScriptFileForShellBleed() parser, which fails to recognize obfuscated command structures; no public exploit code has been identified at time of analysis, though a vendor patch is available.
Denial of service in Rack versions prior to 2.2.23, 3.1.21, and 3.2.6 allows unauthenticated remote attackers to consume disproportionate CPU, memory, I/O, and bandwidth by supplying many small overlapping byte ranges in HTTP Range headers, bypassing the existing CVE-2024-26141 fix that only validates total byte coverage. The vulnerability affects Rack's file-serving paths that process multipart byte range responses, enabling attackers to degrade service availability with minimal request complexity.
Denial of service in Rack::Utils.select_best_encoding allows unauthenticated remote attackers to consume disproportionate CPU resources via a crafted Accept-Encoding header containing multiple wildcard entries, affecting Rack versions prior to 2.2.23, 3.1.21, and 3.2.6. The vulnerability exploits quadratic time complexity in the encoding selection algorithm used by Rack::Deflater middleware, enabling a single HTTP request to trigger sustained CPU exhaustion and application unavailability.
User enumeration in osTicket v1.18.2's password reset endpoint (/pwreset.php) enables remote attackers to discover valid usernames through response analysis, facilitating targeted account compromise attempts. No CVSS score, CISA KEV status, or confirmed patch information is available; exploitation likelihood depends on whether timing or behavioral differences between valid and invalid usernames can be reliably detected without authentication.
Keycloak's SingleUseObjectProvider lacks proper type and namespace isolation, allowing unauthenticated remote attackers with user interaction to delete arbitrary single-use entries and replay consumed action tokens such as password reset links, leading to account compromise. The vulnerability requires user interaction (UI:R) and high attack complexity (AC:H), resulting in a CVSS score of 5.3. No public exploit code or active exploitation has been confirmed at time of analysis.
Rack::Static fails to apply security-relevant response headers to URL-encoded variants of static file paths, allowing attackers to bypass header-based security controls by requesting percent-encoded forms of protected resources. This affects Rack versions prior to 2.2.23, 3.1.21, and 3.2.6, and is particularly dangerous in deployments relying on Rack::Static to enforce Content-Security-Policy, X-Frame-Options, or similar protective headers on static content. No public exploit code or active exploitation has been confirmed at time of analysis.
Unauthenticated remote attackers can access configuration files containing database credentials in MB Connect Line mbconnect24 and mymbconnect24 products, resulting in disclosure of sensitive authentication material. Although CVSS rates this as 5.3 (low severity confidentiality impact), the practical risk is limited because the disclosed credentials cannot be directly exploited to compromise additional systems-no exposed endpoint exists to leverage them. No public exploit code or active exploitation has been identified at the time of analysis.
Rack web server interface versions prior to 2.2.23, 3.1.21, and 3.2.6 fail to properly escape regex metacharacters when constructing directory path filtering expressions, causing the Rack::Directory component to expose full filesystem paths in HTML directory listings. An unauthenticated remote attacker can retrieve sensitive path information by requesting directory listings when the configured root path contains regex special characters such as +, *, or ., achieving low-confidentiality impact with CVSS 5.3. No public exploit code or active exploitation has been confirmed at time of analysis.
SEPPmail Secure Email Gateway before version 15.0.3 allows unauthenticated attackers to inject arbitrary HTML into notification emails about new CA certificates, enabling stored cross-site scripting (XSS) attacks against email recipients. An attacker with the ability to trigger CA certificate notifications can craft malicious HTML payloads that execute when recipients view the notification email, potentially leading to credential theft, malware distribution, or further compromise of email infrastructure. No public exploit code or active exploitation has been confirmed at the time of analysis.
Stored cross-site scripting (XSS) in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the remark parameter in the VPN authentication user management interface (/manage/vpnauthentication/user/). The injected payload persists in the database and executes when other authenticated users access the affected page, enabling session hijacking, credential theft, or lateral privilege escalation within the firewall management console.
Stored cross-site scripting (XSS) in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the remark parameter in the /manage/password/web/ endpoint. The injected payload is persistently stored and executes when other authenticated users access the affected page, enabling session hijacking, credential theft, or lateral movement within the firewall management interface. No public exploit code or active exploitation has been confirmed at the time of analysis.
Stored cross-site scripting in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject malicious JavaScript via the remark parameter in the IPSec management interface (/manage/ipsec/), which persists and executes when other users access the affected page. This requires user interaction (page view) and only affects session integrity and information disclosure within the administrative interface. No public exploit code or active exploitation has been confirmed at the time of analysis.
Stored cross-site scripting (XSS) in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject malicious JavaScript via the remark parameter in the DNS local domains management interface (/manage/dnsmasq/localdomains/). The injected payload persists in the application and executes when other authenticated users access the affected page, enabling session hijacking, credential theft, or malware distribution within the firewall management environment. No public exploit code or active exploitation has been reported.
Stored cross-site scripting (XSS) in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the ADDRESS BCC parameter in /cgi-bin/smtprouting.cgi, with the payload executed when other users view the affected page. The vulnerability requires user interaction (UI:P) and authenticated access (PR:L), limiting real-world impact to internal threat actors with valid credentials, though successful exploitation could compromise session integrity and user data within the firewall management interface.
Stored cross-site scripting in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the domain parameter in the /manage/smtpscan/domainrouting/ endpoint, with execution when other authenticated users view the page. The vulnerability requires user interaction (page view) and authenticated access, resulting in a CVSS score of 5.1 with scope change and integrity impact to other users' sessions. No public exploit code or active exploitation has been identified at time of analysis.
Stored cross-site scripting in Endian Firewall versions 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the DOMAIN parameter in /cgi-bin/smtpdomains.cgi, which is executed when other users view the affected page. The vulnerability requires user interaction and authenticated access but can impact session security and administrative controls. No public exploit code or active exploitation has been identified at time of analysis.
Stored cross-site scripting in Endian Firewall version 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the group parameter in /cgi-bin/proxygroup.cgi, with the malicious payload persisting and executing when other users access the affected page. CVSS score of 5.1 reflects moderate severity with limited scope of impact; exploitation requires prior authentication and user interaction but can affect confidentiality and integrity within the application context. No public exploit code or active exploitation has been confirmed at the time of analysis.
Stored cross-site scripting (XSS) in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the user parameter in /cgi-bin/proxyuser.cgi, which is then executed when other users view the affected page. This requires user interaction (page view) but enables session hijacking, credential theft, or administrative action abuse within the firewall's web interface. No public exploit code or active exploitation has been confirmed at time of analysis.
Stored cross-site scripting (XSS) in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the mimetypes parameter in /cgi-bin/proxypolicy.cgi, which is executed when other users access the affected page. CVSS 5.1 reflects moderate impact; exploitation requires prior authentication and user interaction, limiting real-world severity despite the persistent nature of stored XSS. No public exploit code or active exploitation has been identified at time of analysis.
Stored cross-site scripting (XSS) in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the remark parameter in /cgi-bin/xtaccess.cgi, which is executed when other users view the affected page. The vulnerability requires valid user credentials and user interaction but can compromise session tokens and sensitive data of administrators and other firewall users. No public exploit code or active exploitation has been confirmed at this time.
Stored cross-site scripting (XSS) in Endian Firewall 3.3.25 and earlier allows authenticated attackers to inject arbitrary JavaScript via the remark parameter in /cgi-bin/vpnfw.cgi, which persists and executes when other users access the affected page. CVSS 5.1 reflects low immediate confidentiality/integrity impact but user interaction requirement; the vulnerability requires authenticated access (PR:L), limiting blast radius compared to unauthenticated XSS.
Stored cross-site scripting in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject malicious JavaScript via the remark parameter in /cgi-bin/zonefw.cgi, which persists and executes when other administrators or users access the affected configuration page. The vulnerability requires user interaction (UI:P) and authenticated access (PR:L), limiting immediate risk but enabling account compromise and lateral movement within firewall administrative interfaces.
Endian Firewall 3.3.25 and earlier allows authenticated users to store arbitrary JavaScript in the remark parameter of /cgi-bin/outgoingfw.cgi, which executes when other users view the affected page. This stored cross-site scripting (XSS) vulnerability requires valid login credentials but can compromise session tokens, steal administrative actions, or perform lateral attacks within the firewall management interface. No public exploit code or active exploitation has been confirmed at time of analysis.
Stored cross-site scripting (XSS) in Endian Firewall 3.3.25 and earlier allows authenticated attackers to inject arbitrary JavaScript through the remark parameter in /cgi-bin/incoming.cgi, which is then executed when other users access the affected page. The vulnerability requires user interaction (UI:P) and authenticated access (PR:L), resulting in a CVSS 5.1 score with limited scope impact; no public exploit code or active exploitation has been confirmed.
Stored XSS in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject malicious JavaScript via the remark parameter in /cgi-bin/snat.cgi, which persists and executes when other administrators or users access the affected page. The vulnerability requires low-privilege authentication and user interaction (page view), limiting immediate impact but creating persistent data integrity and session hijacking risks within the appliance administrative interface.
Stored cross-site scripting in Endian Firewall 3.3.25 and prior allows authenticated users to inject malicious JavaScript via the remark parameter in /cgi-bin/dnat.cgi, which persists and executes when other administrators or users access the affected page. This requires valid login credentials but can compromise the integrity and confidentiality of management sessions for other users.
Stored cross-site scripting in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the dscp parameter in the QoS rules management interface (/manage/qos/rules/). When other authenticated users view the affected configuration page, the injected script executes in their browser context, enabling session hijacking, credential theft, or lateral movement within the firewall management console. EPSS risk is elevated at moderate severity (CVSS 5.1), and no public exploit code or active exploitation has been confirmed.
Stored cross-site scripting in Endian Firewall 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the name parameter in the QoS classes management interface (/manage/qos/classes/), which is executed when other users access the affected page. The vulnerability requires user interaction and authentication, resulting in a CVSS 5.1 score with limited scope of impact; no public exploit code or active exploitation has been confirmed.
Stored cross-site scripting (XSS) in Endian Firewall version 3.3.25 and prior allows authenticated attackers to inject arbitrary JavaScript via the remark user ham spam parameter in /cgi-bin/salearn.cgi. The injected payload is stored and executed in the browsers of other users who view the affected page, enabling session hijacking, credential theft, or malware distribution. No public exploit code or active exploitation has been confirmed; real-world risk is limited by the authentication requirement and user interaction dependency.
Endian Firewall 3.3.25 and earlier contains stored cross-site scripting (XSS) in the DHCP fixed leases management interface, where the remark parameter fails to sanitize user input. An authenticated attacker can inject malicious JavaScript into the remark field at /manage/dhcp/fixed_leases/ that persists in the application and executes in the browsers of other administrators viewing the same page, enabling session hijacking, credential theft, or unauthorized configuration changes. No public exploit code or active exploitation has been confirmed; however, the vulnerability requires only low-privilege authentication and normal user interaction to trigger.