CVE-2026-97693
Description détaillée
Yuxi is a large-model-based intelligent knowledge base and knowledge graph agent development platform. From 0.6.0.beta1 until 0.7.1, there is insufficient isolation between execution-capable Docker sandboxes and trusted application services. A low-privileged authenticated user who can execute commands in their own sandbox can reach other tenants' sandbox APIs, access their files and secrets, execute commands in active sandboxes, and enumerate, create, or delete sandboxes through the unauthenticated provisioner without knowing administrator, database, or MinIO credentials. Sandboxes can also reach shared internal services; if shipped default service credentials remain unchanged, an attacker can use them, and if the exposed PostgreSQL account is a superuser, obtain command execution in the PostgreSQL container. These paths can disclose or modify cross-tenant data and sandbox state or cause denial of service through sandbox deletion without a container escape or kernel vulnerability. This vulnerability is fixed in 0.7.1. Existing deployments must separately rotate persisted PostgreSQL, Neo4j, or MinIO credentials that were previously set to documented or shipped defaults because upgrading the Compose files does not automatically rotate them.
Vecteur d'attaque (CVSS)
Dernières Vulnérabilités
CVE-2026-97702
Open edX Platform enables the authoring and delivery of online learning at any scale. From the Ulmo release until release/ulmo.4 and release/verawood.1, the GET /api/support/v1/manage_course_team/ endpoint in CourseTeamManageAPIView requires only IsAuthenticated and omits the global support-role authorization check intended for this help-desk surface. Any authenticated learner can supply an email, username, or user_id for an arbitrary target and use the 200 and 404 response discrepancy to distinguish an active account from an absent or inactive account. A caller with a course-level or organization-level instructor role can additionally obtain the target user's staff or instructor assignments within the caller's course or organization scope. Anonymous callers remain blocked, and the separately authorized PUT behavior is outside this vulnerability. This issue is fixed in release/ulmo.4 and release/verawood.1.
CVE-2026-86041
ezBookkeeping is an open source, self-hosted personal finance application. From 0.10.0 until 1.6.0, ezBookkeeping constructs its Gin router with gin.New() without configuring SetTrustedProxies, causing c.ClientIP() to trust a spoofed X-Forwarded-For value. An unauthenticated remote attacker can supply an allowlisted address to bypass the mcp_allowed_remote_ips or api_token_allowed_remote_ips gate. A valid MCP or API token remains required for protected actions. When such actions occur, the same spoofed value is stored in Transaction.CreatedIp and the request log, attributing activity to the trusted host instead of the attacker. This issue is fixed in version 1.6.0.
CVE-2026-77372
VeraCrypt provides disk encryption with strong security based on TrueCrypt. Prior to 1.26.30, the Unix, macOS, and FreeBSD serialization paths in src/Platform/Serializer.cpp and src/Volume/VolumePassword.cpp accept untrusted length and count values when deserializing data that can cross from the unprivileged GUI or CLI to the root CoreService helper. Serializer::DeserializeWString allocates floor(size / sizeof(wchar_t)) elements but reads size bytes, so a non-aligned size can overflow the heap allocation. VolumePassword::Deserialize accepts PasswordSize values larger than the fixed 128-byte PasswordBuffer, allowing downstream DataPtr() and Size() consumers such as the KDF to read beyond the buffer. DeserializeString, DeserializeWString, and the string and wide-string list readers also lack allocation and collection bounds, which can exhaust memory, and a zero-length DeserializeString can access an empty vector. Windows uses a different path, and the immediate CoreService pipe writer is the same-user parent process, so the documented exposure is a local defense-in-depth boundary rather than a remote attack. This issue is fixed in 1.26.30.