Veille CVE

Vendredi 21 août 2026

9 CVE

Critique 1 Élevée 5 Moyenne 3 Faible 0 Non évaluée 0 Exploitées (KEV) Tri : plus récentes « git » ✕
  1. CVE-2026-44517 6.3 Moyenne

    containers buildah

    Buildah is a tool that facilitates building OCI images. From 1.38.1 until 1.43.2 and 1.44.0, TempDirForURL in define/types.go does not securely confine Git repository subdirectories to the downloaded build context, and downloadToDirectory and stdinToDirectory can follow a Dockerfile symlink left by a partially extracted tar archive. A malicious server supplying a Git repository or tar archive can cause files outside the build context directory to be included in the context or copied into the build. This issue is fixed in versions 1.43.2 and 1.44.0.

  2. CVE-2026-62960 7.4 Élevée

    git-for-windows git

    Git for Windows is the Windows port of Git. Prior to 2.55.0.windows.4, a malicious remote Git server can advertise a bundle URI that reaches transport_get_remote_bundle_uri(), fetch_bundle_uri_internal(), and copy_uri_to_file() in bundle-uri.c during clone or fetch when transfer.bundleuri=true. Non-HTTP(S) values are treated as local filesystem paths, and file URI prefixes are removed, so a bare UNC path or file URI targeting an attacker-controlled share causes Windows to initiate an outbound SMB connection. This can expose NTLM authentication material to the attacker-selected host. This issue is fixed in version 2.55.0.windows.4.

  3. CVE-2026-62676 7.1 Élevée

    omnigent-ai omnigent

    Omnigent is an open-source AI agent framework and meta-harness for orchestrating coding agents. Prior to 0.3.0, the shared shell-command parser in omnigent/policies/builtins/_shell.py fails to recognize combined interpreter flags, the timeout, nice, setsid, and stdbuf wrappers, command substitutions, and a single background control operator. A gated git push or gh write hidden with these forms produces no parsed operation, causing the github.py write_repos and write_branches allowlist and the working_dir.py workspace confinement policies to abstain and allow the command. An authenticated or prompt-injected agent can therefore push to an unauthorized repository or branch or escape the intended workspace. This issue is fixed in version 0.3.0.

  4. CVE-2026-77812 9.4 Critique

    DJI Neo · DJI Neo 2 · DJI Flip · DJI Air 3

    DJI drones transmit DUML (DJI Universal Markup Language) protocol messages over BLE (Bluetooth Low Energy) without encryption. When a client attempts to connect to the drone over Wi-Fi, or when the drone is switched to QuickTransfer mode, the DJI Fly application exchanges DUML messages with the drone over BLE, including the Wi-Fi credentials. An attacker within BLE range can passively sniff this traffic and recover the credentials in cleartext, including the drone's Wi-Fi PSK, SSID, and trusted identifier UUID. Obtaining these credentials allows the attacker to join the drone's internal Wi-Fi network, interact with network services exposed by the drone, and decrypt Wi-Fi traffic exchanged between the drone and the legitimate user. * An attacker within BLE range recovers the Wi-Fi SSID and PSK in cleartext, and can then join the drone's network * The same capture also exposes the session UUID identifier, which is the only thing the drone uses to tell a trusted client from an unknown one, so the attacker can replay it and skip the physical confirmation of new connected devices. * The credentials do not change between sessions unless the operator manually resets the Wi-Fi settings, so one capture stays valid indefinitely * The attack is fully passive, with nothing transmitted and no connection made, so neither the operator nor the drone has any indication the session was observed * A BLE sniffer and presence during one normal DJI Fly connection are needed Affected models are DJI Neo until 01.00.0400, DJI Neo 2 until 01.00.0500, DJI Flip until 01.00.1200, DJI Air 3 until 01.00.1600, DJI Air 3S until 01.00.1400, DJI Avata 2 until 01.00.0400, DJI Avata 360 until 01.00.0300, DJI Mavic 3 until 01.00.1400, DJI Mavic 3 Classic until 01.00.0800, DJI Mavic 3 Pro until 01.01.0700, DJI Mavic 4 Pro until 01.00.0500, DJI Mini 2 until 01.07.0200, DJI Mini 3 until 01.00.0500, DJI Mini 3 Pro until 01.00.0900, DJI Mini 4 Pro until 01.00.1100, and DJI Mini 5 Pro until 01.00.0600. Remediation requires a firmware update from the vendor. There is no user-side mitigation that fully addresses the vulnerability without upgrading.

  5. CVE-2026-77780 5.3 Moyenne

    Roskus Prospero Flow CRM

    Authorization Bypass Through User-Controlled Key in the transaction save endpoint in Roskus Prospero Flow CRM 4.9.1 through 5.14.0 allows a user with transaction and accounting creation permissions to disclose another company's bank account name, bank name and card last four digits via a bank_account_id or bank_card_id belonging to that company in POST /transaction/save, which is persisted and rendered without any company ownership check.

  6. CVE-2026-59279 7.5 Élevée

    Spring AI · Vmware spring ai

    The MCP Streamable HTTP server transport (WebFlux and WebMvc variants) does not place any limit on the number of sessions it retains, and by default does not require clients to be authenticated. As a result, a remote attacker can cause the server to accumulate an unbounded number of sessions over time, gradually exhausting available memory and ultimately causing a Denial of Service that affects all legitimate clients. Affected versions: Spring AI: 2.0.0

  7. CVE-2026-77751 8.8 Élevée

    misp-stix

    A path traversal vulnerability existed in the handling of MISP object template names during STIX 2 import and MISP-to-STIX 2 export. MISP object names are passed to PyMISP's object-template resolution mechanism, which constructs a filesystem path by joining the configured MISP object-template directory, the object name, and definition.json. An object name originating from untrusted STIX or MISP content was not sufficiently restricted before being used in this filesystem path. An attacker able to supply a crafted object name containing path separators or traversal sequences such as ../ could therefore cause template resolution to escape the expected template directory and attempt to load a definition.json file from another location accessible to the process. During STIX 2 import, an attacker-controlled x_misp_name from a custom STIX object could directly reach this template-resolution mechanism. The issue could also become persistent. A malicious object name stored in a MISP event could later be processed again during STIX 2 export. Consequently, content originally introduced in one security context could trigger filesystem access later when the event is exported by a process operating with different or greater privileges. If a suitable definition.json file exists outside the intended template directory, its contents may be interpreted as a MISP object template and fields from that file copied into the converted object. This can result in unintended disclosure of locally accessible data represented by the template file and modification of the resulting object's metadata or semantics. The patches introduce strict validation of object-template names. Valid names are restricted to a single path component containing letters, digits, hyphens, or underscores. Names that do not meet these requirements are replaced with the generic unknown-template name before reaching PyMISP template resolution. The original rejected name is preserved in the object's comment and a warning is generated, preventing traversal while retaining the source information.

  8. CVE-2026-63046 8.8 Élevée

    Apache Software Foundation Apache InLong · Apache inlong

    Improper Neutralization of Argument Delimiters in a Command ('Argument Injection') vulnerability in Apache InLong. Agent Installer's ModuleManager executes arbitrary shell commands via ExcuteLinux.exeCmd() with no filtering or whitelist validation. This issue affects Apache InLong: from 2.0.0 before 2.4.0. Users are advised to upgrade to Apache InLong's 2.4.0 or cherry-pick [1]/[2] to solve it. [1]  https://github.com/apache/inlong/pull/12151 . [2]  https://github.com/apache/inlong/pull/12155 .

  9. CVE-2026-76131 5.3 Moyenne

    Yamaha Corporation VOCALOID6

    Use of hard-coded credentials issue exists in VOCALOID6 , which may allow an attacker to impersonate a legitimate VOCALOID6 Editor and gain access to Yamaha's activation and content servers.