CVEs

cJSON versions 1.5.0 through 1.7.19 contain an incorrectly-resolved name or reference vulnerability in the decode_pointer_inplace() function within cJSON_Utils.c that allows unauthenticated attackers to cause JSON Patch operations to target wrong object keys by supplying crafted JSON Pointer escape sequences (~0 or ~1) in patch paths. Attackers can submit malicious RFC 6902 JSON Patch input to applications using cJSONUtils_ApplyPatches() or cJSONUtils_ApplyPatchesCaseSensitive() to silently corrupt data or delete unintended keys, potentially bypassing authorization controls in applications that rely on JSON Patch for access-controlled data modification.

Use after free in Blink in Google Chrome prior to 151.0.7922.137 allowed a remote attacker to execute arbitrary code inside a sandbox via a crafted HTML page. (Chromium security severity: High)

Use after free in HTML in Google Chrome prior to 151.0.7922.137 allowed a remote attacker to execute arbitrary code inside a sandbox via a crafted HTML page. (Chromium security severity: High)

Use after free in Extensions in Google Chrome prior to 151.0.7922.137 allowed an attacker who convinced a user to install a malicious extension to execute arbitrary code inside a sandbox via a crafted Chrome Extension. (Chromium security severity: High)

Use after free in TabStrip in Google Chrome on Mac prior to 151.0.7922.137 allowed a remote attacker who had compromised the renderer process to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High)

Use after free in V8 in Google Chrome prior to 151.0.7922.137 allowed a remote attacker to execute arbitrary code inside a sandbox via a crafted HTML page. (Chromium security severity: High)

A MongoDB driver component could write sensitive configuration information, including a credential used for outbound network connectivity, to application log output in cleartext during routine client initialization. This occurs automatically as part of normal operation and requires no special privileges to trigger. A party able to read the affected application's logs or downstream log-aggregation storage could recover the credential and reuse it to authenticate to the associated network infrastructure. This issue affects confidentiality only.

calibre is an e-book manager. Prior to 9.12.0, calibre processes attacker-controlled composite_template metadata from a malicious EPUB, OPF, PDF, or similar file through program: and a nested template() call whose formatter does not inherit allow_python_templates=False, allowing a nested python: template to reach compile_python_template and execute arbitrary Python code when the file is opened or imported. This issue is fixed in version 9.12.0.

Kestra is an open-source, event-driven orchestration platform. Prior to 2.0.0, Kestra's core/src/main/java/io/kestra/core/runners/pebble/functions/HttpFunction.java passes the user-controlled http() uri argument to URI.create() and the server-side HTTP client without restricting private, loopback, or link-local destinations, allowing an unauthenticated attacker to import and execute a flow that accesses internal services or cloud metadata.

libgit2 versions v0.27.0 through v1.9.0 built with the libssh2 SSH backend (USE_SSH=libssh2) contain a shell command injection vulnerability that allows remote attackers to execute arbitrary commands on an SSH server by supplying a repository path containing unescaped shell metacharacters such as single quotes, semicolons, or pipes. The gen_proto() function in ssh_libssh2.c inserts the repository path directly into a shell command string without escaping special characters before passing it to libssh2_channel_exec(), enabling an attacker to craft a malicious submodule URL in a .gitmodules file that, when processed during a recursive clone, causes the remote server's shell to interpret injected commands under the victim's SSH user account.

In the Mira hormone monitor device firmware v1.7.1.47 build 01070147, a remote unauthenticated attacker within BLE range (approximately 10–30 meters) can silently rebind the device to an attacker-controlled account, extract stored hormone measurements in cleartext, cause a denial-of-service via malformed or undocumented command opcodes, and passively track the user via a static random BLE address that never rotates.

The distributed Mira Android APK v4.5.15.4 allows an attacker read/write access to reproductive health profiles from internet connected hosts, which could result in forgery, deletion, or destruction of health information.

The login endpoint on the Mira cloud API accepts any format-valid string in the password field and returns a live active session token for the account matching the supplied email address. An attacker could use an email address to control cloud accounts and access hormone record information and account settings.

Improper TLS hostname verification vulnerability in Apache HttpComponents Client 5.4 or newer. HostnameVerificationPolicy#BUILTIN setting has no effect when used with the async version of HttpClient. An attacker that can intercept and modify traffic between the client and the server can impersonate the server by presenting a valid certificate for a different domain. Please note the classic version of HttpClient is not affected by this vulnerability. Affected users are recommended to upgrade to at least version 5.6.4, which fixes the issue.

When the Mira Android app opens in-app WebView content (e.g., shop redirect flows), the user's live session token is appended to the URL as a query string parameter, and a persistent user identifier is included in the WebView's User-Agent header. Both are then transmitted to third-party web properties, referrer logs, and any JavaScript running in the WebView context.

An insufficient certificate validation in a privileged communication workflow, was identified in a GMS application 9.5.1 (Build 9510.1044) and earlier versions which, under a successful MitM attack and controlled network conditions, could permit unauthorized changes.

Improper Control of Generation of Code ('Code Injection') Vulnerability in the SonicWall Email Security appliance allows an authenticated attacker with access to the SonicWall Email Security restricted CLI can inject arbitrary OS commands that execute as root via SNMP.

Improper Control of Generation of Code ('Code Injection') Vulnerability in the SonicWall Email Security appliance allows an authenticated attacker with access to the SonicWall Email Security restricted CLI can inject arbitrary OS commands that execute as root via netmask.

An authenticated command injection vulnerability was identified in GMS Command-Line Interface (CLI) 9.5.1 (Build 9510.1044) and earlier versions which allows low-privileged local user to execute system commands with root privileges.

An unauthenticated command injection vulnerability was identified in the GMS Dispatcher Service in GMS 9.5.1 and earlier versions which allows remote attacker to perform remote code execution through specially crafted requests.

Malcolm is a network traffic analysis tool suite. Prior to version 26.07.0, role-based access control enforced in the Nginx OpenResty Lua layer evaluates the raw, unnormalized `ngx.var.request_uri`, while Nginx itself routes requests using the normalized path. An authenticated low-privilege user can prepend a traversal segment (for example `/x/../upload/...`) so that Nginx routes the request to a restricted backend while the Lua role check fails to match any rule and falls open, granting access it should deny. Version 26.07.0 fixes the issue.

Malcolm is a network traffic analysis tool suite. Prior to version 26.07.0, `safe-extract.py` protects file extraction with libarchive's secure flags, but creates directory entries with a raw `os.makedirs(os.path.join(dest, entry.pathname))` that has no traversal protection. An uploaded malicious archive containing a directory entry with a `../` sequence or an absolute path causes the filebeat processing container to create directories outside the intended extraction directory. Version 26.07.0 fixes the issue.

Malcolm is a network traffic analysis tool suite. Prior to version 26.07.0, `safe-extract.py` extracts uploaded archives with no limit on entry count, directory depth, total entries, or output size. A small malicious archive containing a large number of directory or file entries causes the filebeat processing container to create an unbounded number of filesystem objects, exhausting inodes or filesystem metadata and denying service to the processing pipeline and any service sharing the same mount. Version 26.07.0 fixes the issue.

Malcolm is a network traffic analysis tool suite. The file-upload component (FilePond PHP backend) accepts uploads at `POST /server/php/submit.php` and stores them in a directory served by the same nginx and php-fpm instance. The allow-list that should restrict accepted file types is an empty array by default (`file-upload/php/config.php:16`), so the type check is a no-op and every extension is accepted. The filename sanitizer keeps the `.php` extension intact. Committed files land in `/var/www/upload/server/php/files` (`file-upload/php/config.php:7`), and the component's nginx routes any URL ending in `.php` to php-fpm. An authenticated `GET /server/php/files/<name>.php` then executes the uploaded code as `www-data`. Prior to version 26.06.1, in RBAC mode, the upload endpoint is reachable by the granular `ROLE_UPLOAD` role (`nginx/lua/nginx_auth_helpers.lua:71`), a role intended only for submitting capture files. As a result, a user holding the upload-only role runs arbitrary PHP as `www-data` inside the file-upload container. Version 26.06.1 fixes the issue.

TypeBot is a chatbot builder tool. Versions prior to 3.17.0 allow a low-privilege read collaborator to extract a workspace OAuth `credentialsId` from a readable bot configuration and then overwrite that credential through `handleUpdateOAuthCredentials()` by supplying an attacker-controlled writable `workspaceId`. The update path validates only the attacker-supplied workspace and then updates the credential record by global `id` alone, while also rewriting the credential's `workspaceId`. This allows cross-workspace OAuth credential takeover and reassignment. Version 3.17.0 patches the issue.

TypeBot is a chatbot builder tool. Versions prior to 3.17.0 expose a deprecated public upload endpoint at `GET /api/v1/typebots/{typebotId}/blocks/{blockId}/storage/upload-url` that accepts an attacker-controlled `filePath` and returns a presigned S3 `PUT` URL for that exact key. Because the endpoint only checks that the referenced typebot is public and that the referenced block is a file input block, an unauthenticated attacker who knows a valid public `typebotId` and `blockId` can request presigned upload URLs for arbitrary objects in the shared bucket, including `private/...` and other tenants' `public/...` paths. Version 3.17.0 fixes this issue.

TypeBot is a chatbot builder tool. Prior to version 3.16.0, the OpenAI "Create Transcription" action handler fetches a user-supplied audio URL using `fetch()` without applying the SSRF protection that exists elsewhere in the codebase. An attacker can direct the server to make HTTP requests to arbitrary internal addresses and localhost. The fetched content is passed to the OpenAI Whisper API and the transcription result is returned to the attacker. Version 3.16.0 fixes the issue.

CivetWeb (commit 4a4f0c95) contains a heap and stack buffer overflow vulnerability in the read_websocket() function that allows unauthenticated remote attackers to corrupt memory by sending compressed WebSocket frames when both USE_ZLIB and MG_EXPERIMENTAL_INTERFACES are defined. Attackers can negotiate permessage-deflate during the WebSocket handshake and send a crafted frame with the RSV1 bit set, causing the server to write a 4-byte zlib sync trailer out-of-bounds past the allocated buffer, leading to heap metadata corruption, denial of service, or potential code execution.

Snipe-IT before 8.6.0 contains an authorization bypass (insecure direct object reference) in the asset checkout-request cancellation endpoint. The cancel_by_admin and requestingUser values are read from user-controlled URL path segments and used without a server-side authorization check, so any authenticated, low-privileged user can supply a non-empty cancel_by_admin value to bypass the request-ownership check and cancel another user's pending checkout request. Because asset and user identifiers are sequential integers, an attacker can enumerate them to cancel every pending checkout request, disrupting the asset-request workflow. This is fixed in Snipe-IT 8.6.0.

A flaw was found in FreeIPA. The trust-fetch-domains command is gated by a read-only permission on the trust object rather than a trust-administration permission, allowing an authenticated, non-privileged IPA user to trigger a privileged Active Directory trust refresh using an attacker-supplied server and credentials, resulting in unauthorized, attacker-controlled modification of trusted-domain and ID-range identity data in the IPA LDAP directory.

An insecure handling of serialized objects vulnerability was found in the one of the service of GMS application 9.5.1 (Build 9510.1044) and earlier versions. A local attacker with the ability to interact with the service could exploit this behavior to perform unauthorized actions through the affected component.

The Frontend Admin by DynamiApps plugin for WordPress is vulnerable to authorization bypass in all versions up to, and including, 3.29.9. This is due to the plugin not properly verifying that a user is authorized to perform an action. This makes it possible for authenticated attackers, with subscriber-level and above permissions, to reset the password of any user on the site, including administrators, leading to full account takeover and complete site compromise. Exploitation requires the attacker to hold a valid encrypted Current-User token obtained by accessing any Edit User form they are legitimately authorized to submit, which they then use as a known-plaintext base for the CBC bit-flipping forgery.

FileRun up to and including version 2026.2.0 contains an OS command injection vulnerability that allows authenticated attackers to achieve remote code execution by uploading a file with a malicious filename containing shell command substitution sequences. The thumbnail generation system passes filenames wrapped in shell double-quotes directly to exec() without escapeshellarg() sanitization, allowing filenames such as $(PAYLOAD).mp4 to survive the filename sanitizer and be evaluated as shell commands when ffmpeg, ImageMagick, vips, or stl-thumb processes the file during thumbnail generation.

In sshd in OpenSSH before 10.5, the restrict keyword (in authorized_keys) was supposed to be applicable to tunnel forwarding but was not.

In ssh in OpenSSH before 10.5, a use-after-free for realloc data can occur if a certain pair of remote-forwarding operations are concurrent.

In ssh-agent in OpenSSH before 10.5, some operations can occur remotely but were intended to occur only locally, including operations that add tokens or use keys. This is caused by misinteraction between agent locking and the [email protected] extension.

kkFileView is a universal file online preview project based on Spring Boot. Prior to 5.0.1, the unauthenticated POST /listFiles endpoint in server/src/main/java/cn/keking/web/controller/FileController.java passes the user-controlled path parameter from FileController#getFiles to Files.newDirectoryStream without confinement to the demo directory, allowing directory enumeration outside the intended root. This issue is fixed in version 5.0.1.

kkFileView is a universal file online preview project based on Spring Boot. Prior to 5.0.1, the unauthenticated GET /addTask endpoint in kkFileView is omitted from TrustHostFilter and TrustDirFilter in server/src/main/java/cn/keking/config/WebConfig.java, allowing FileConvertQueueTask to fetch an attacker-selected URL after FileHandlerService#getFileAttribute uses the fullfilename parameter to force an OFFICE, COMPRESS, or CAD type. This issue is fixed in version 5.0.1.

FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to 3.30.0, FreeRDP's winpr/libwinpr/sspi/Kerberos/kerberos.c kerberos_DecryptMessage function fails to bound the peer-controlled GSS Wrap-token EC field before using it with RRC in IOV pointer offsets, allowing a malicious RDP peer to trigger out-of-bounds reads and in-place writes during CredSSP/NLA Kerberos decryption. This issue is fixed in version 3.30.0.

FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to 3.30.0, FreeRDP server-side RDSTLS in libfreerdp/core/rdstls.c accepts an attacker-supplied RDSTLS_TYPE_CAPABILITIES PDU while rdstls_server_authenticate is waiting for RDSTLS_TYPE_AUTHREQ, leaving resultCode at RDSTLS_RESULT_SUCCESS and allowing a remote unauthenticated client to bypass the RedirectionGuid, username, domain, or password checks. This issue is fixed in version 3.30.0.

FreeCAD is a free and open-source multiplatform 3D parametric modeler. Prior to 1.1.2, the Xerces SAX2 XMLReader constructed in src/Base/Reader.cpp by Base::XMLReader::XMLReader() parses attacker-controlled Document.xml from a crafted .FCStd archive without disabling default external entity resolution or external DTD loading. When Document::restore() opens the document, external entities can read local files through the file URI scheme or initiate server-side requests through the http URI scheme, and resolved content can flow through the characters() callback. This issue is fixed in version 1.1.2.

FreeCAD is a free and open-source multiplatform 3D parametric modeler. Prior to 1.1.2, PropertyFileIncluded::Restore() in src/App/PropertyFile.cpp concatenates an attacker-controlled file or data attribute from Document.xml with the document transient path without rejecting directory components, absolute paths, or parent traversal. A crafted .FCStd archive with a matching FileIncluded XML attribute and ZIP entry can therefore write attacker-controlled content to arbitrary locations accessible to the FreeCAD user, potentially enabling persistence, credential compromise, configuration replacement, or code execution. This issue is fixed in version 1.1.2.

FreeCAD is a free and open-source multiplatform 3D parametric modeler. Prior to 1.1.2, the FEM Displacement Constraint task dialog in src/Mod/Fem/Gui/TaskFemConstraintDisplacement.cpp passes the xDisplacementFormula, yDisplacementFormula, and zDisplacementFormula fields of a Fem::ConstraintDisplacement object through TaskDlgFemConstraintDisplacement::accept() into Gui::Command::doCommand. The escaping helper neutralizes quotation marks but not backslashes, allowing crafted formula text to terminate the generated Python string and execute arbitrary Python code with the FreeCAD process's privileges when a victim accepts the dialog. This issue is fixed in version 1.1.2.

ffuf is a fast web fuzzer written in Go. Prior to 2.2.0, ffuf allows a malicious target server to cause an out-of-memory denial of service because the response size guard in pkg/runner/simple.go checks only the compressed Content-Length while io.ReadAll reads gzip, brotli, deflate, transparently decompressed, or chunked response bodies without a decompressed-size bound. This issue is fixed in version 2.2.0.

Faker generates massive amounts of fake data in the browser and Node.js. Prior to 10.5.0, the faker.helpers.fake method in src/modules/helpers/eval.ts allows attacker-controlled fake templates to access the Function constructor through fakeEval.resolveProperty when a function returns another function, enabling arbitrary JavaScript code execution. This issue is fixed in version 10.5.0.

Ente provides end-to-end encrypted cloud services and security tools. Prior to 2026.07.28, Ente 2of3 card format version 1 stored the secret byte length and 32-bit FNV-1a checksum in cleartext on every card, allowing someone with one card to test candidate secrets offline and recover low-entropy or predictable secrets. This issue is fixed in version 2026.07.28.

Django REST framework is a powerful and flexible toolkit for building Web APIs. Prior to 3.17.2, Django REST Framework's rest_framework/renderers.py AdminRenderer.render() uses override_method() to simulate GET and directly invokes view.get() without view.check_permissions() while rendering an invalid write request, allowing a 400 Bad Request HTML response to disclose data from a GET representation that the requester is not permitted to access. This issue is fixed in version 3.17.2.

Bash-it 3.2.0 contains a terminal escape sequence injection vulnerability in the barbuk theme's Python virtualenv prompt segment that allows local attackers to inject arbitrary terminal control sequences by embedding escape sequences in the requires-python field of a pyproject.toml file. When a user navigates into a directory containing a maliciously crafted pyproject.toml, the unfiltered field value is read via awk and concatenated directly into PS1 through __prompt-command without stripping control characters, causing injected OSC or CSI sequences to be written to and interpreted by the terminal emulator on every prompt render.

DB-GPT v0.8.1 contains an unauthenticated path traversal vulnerability that allows remote attackers to write arbitrary files to any location on the server by injecting directory traversal sequences into the user_id HTTP header of the Python file-upload endpoint. Attackers can send a crafted multipart upload request with a traversal-poisoned user_id header to escape the intended upload directory and write attacker-controlled content to locations such as Python startup hooks, cron directories, or agent scripts, resulting in remote code execution.

PapersGPT for Zotero 0.6.1 contains a remote code execution vulnerability that allows attackers to execute arbitrary JavaScript by returning malicious code from an LLM endpoint that is passed unsanitized to window.eval() in views.ts. Attackers can exploit this through prompt injection in PDFs, MITM interception of API requests, or a malicious custom LLM endpoint to execute arbitrary code in Zotero's chrome-privileged context, enabling file read/write, process execution, and access to all Zotero data.

A flaw was found in Keycloak. An authenticated administrator with the `manage-clients` role can exploit a Time-of-check to time-of-use (TOCTOU) vulnerability in the name-based admin role checks. This allows the attacker to escalate their privileges to `realm-admin` for all users within the realm, granting them extensive control over the system. The composite role relationship persists even after the attacker's own permissions are revoked and across system reboots.

Updated: 2026-08-11

DBI versions before 1.648 for Perl saved errors in a limited-sized buffer. Error messages that were returned when RaiseError, PrintError or HandleError were set were written to a 200-byte buffer without a length limit. Attackers that can influence the error text in an application can trigger a buffer overflow.

Updated: 2026-08-11

bz2.BZ2Decompressor objects could be reused after a decompression error. If an application caught the resulting OSError and retried with the same decompressor, crafted input could cause the decompressor to resume from an invalid internal state and perform out-of-bounds writes to a stack buffer. This could crash the process when processing untrusted data.

Updated: 2026-08-11

NGINX Plus and NGINX Open Source have a vulnerability in the ngx_http_rewrite_module module. This vulnerability exists when a rewrite directive uses a regex pattern with distinct, overlapping Perl-Compatible Regular Expression (PCRE) captures (for example, ^/((.*))$) and a replacement string that references multiple such captures (for example, $1$2) in a redirect or arguments context. An unauthenticated attacker along with conditions beyond their control can exploit this vulnerability by sending crafted HTTP requests. This may cause a heap buffer overflow in the NGINX worker process leading to a restart. Additionally, attackers can execute code on systems with Address Space Layout Randomization (ASLR) disabled or when the attacker can bypass ASLR. Note: Software versions which have reached End of Technical Support (EoTS) are not evaluated.

Updated: 2026-08-11

Insufficient input validation vulnerability in the NETGEAR R7000 models allows authenticated administrators connected to the local network to make unauthorized modification to router software and functionality.

Updated: 2026-08-11

Untrusted Pointer Dereference in ASUS GPU Tweak III, GPUTweakII, AI Suite3, and VGAdll: An IOCTL vulnerability allows a local attacker to write a specific value to an arbitrary memory address, potentially leading to privilege escalation. Refer to the ' Security Update for ASUS GPU Tweak III, GPU Tweak II, AI Suite 3, and Armoury Crate Security Bulletin ' section on the ASUS Security Advisory for more information.

Updated: 2026-08-11

IBM Security Verify could allow a remote attacker to obtain sensitive information when a detailed technical error message is returned in the browser. This information could be used in further attacks against the system.

Updated: 2026-08-11

The Signed Video Framework contained a buffer overflow issue which could lead the application using this framework to crash. The issue exclusively affects the tools used for the validation of signed content. The AXIS OS device's signed video functionality remains unaffected.

Updated: 2026-08-11

Incorrect boundary conditions in the Audio/Video: Playback component. This vulnerability was fixed in Firefox 150, Thunderbird 150, Firefox ESR 140.10.1, Thunderbird 140.10.1, and Firefox ESR 115.35.2.

Updated: 2026-08-11

In sshd in OpenSSH before 10.5, the restrict keyword (in authorized_keys) was supposed to be applicable to tunnel forwarding but was not.

Updated: 2026-08-11

In ssh in OpenSSH before 10.5, a use-after-free for realloc data can occur if a certain pair of remote-forwarding operations are concurrent.

Updated: 2026-08-11

In ssh-agent in OpenSSH before 10.5, some operations can occur remotely but were intended to occur only locally, including operations that add tokens or use keys. This is caused by misinteraction between agent locking and the [email protected] extension.

Updated: 2026-08-11

kkFileView is a universal file online preview project based on Spring Boot. Prior to 5.0.1, the unauthenticated POST /listFiles endpoint in server/src/main/java/cn/keking/web/controller/FileController.java passes the user-controlled path parameter from FileController#getFiles to Files.newDirectoryStream without confinement to the demo directory, allowing directory enumeration outside the intended root. This issue is fixed in version 5.0.1.

Updated: 2026-08-11

kkFileView is a universal file online preview project based on Spring Boot. Prior to 5.0.1, the unauthenticated GET /addTask endpoint in kkFileView is omitted from TrustHostFilter and TrustDirFilter in server/src/main/java/cn/keking/config/WebConfig.java, allowing FileConvertQueueTask to fetch an attacker-selected URL after FileHandlerService#getFileAttribute uses the fullfilename parameter to force an OFFICE, COMPRESS, or CAD type. This issue is fixed in version 5.0.1.

Updated: 2026-08-11

FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to 3.30.0, FreeRDP's winpr/libwinpr/sspi/Kerberos/kerberos.c kerberos_DecryptMessage function fails to bound the peer-controlled GSS Wrap-token EC field before using it with RRC in IOV pointer offsets, allowing a malicious RDP peer to trigger out-of-bounds reads and in-place writes during CredSSP/NLA Kerberos decryption. This issue is fixed in version 3.30.0.

Updated: 2026-08-11

FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to 3.30.0, FreeRDP server-side RDSTLS in libfreerdp/core/rdstls.c accepts an attacker-supplied RDSTLS_TYPE_CAPABILITIES PDU while rdstls_server_authenticate is waiting for RDSTLS_TYPE_AUTHREQ, leaving resultCode at RDSTLS_RESULT_SUCCESS and allowing a remote unauthenticated client to bypass the RedirectionGuid, username, domain, or password checks. This issue is fixed in version 3.30.0.

Updated: 2026-08-11

FreeCAD is a free and open-source multiplatform 3D parametric modeler. Prior to 1.1.2, the Xerces SAX2 XMLReader constructed in src/Base/Reader.cpp by Base::XMLReader::XMLReader() parses attacker-controlled Document.xml from a crafted .FCStd archive without disabling default external entity resolution or external DTD loading. When Document::restore() opens the document, external entities can read local files through the file URI scheme or initiate server-side requests through the http URI scheme, and resolved content can flow through the characters() callback. This issue is fixed in version 1.1.2.

Updated: 2026-08-11

FreeCAD is a free and open-source multiplatform 3D parametric modeler. Prior to 1.1.2, PropertyFileIncluded::Restore() in src/App/PropertyFile.cpp concatenates an attacker-controlled file or data attribute from Document.xml with the document transient path without rejecting directory components, absolute paths, or parent traversal. A crafted .FCStd archive with a matching FileIncluded XML attribute and ZIP entry can therefore write attacker-controlled content to arbitrary locations accessible to the FreeCAD user, potentially enabling persistence, credential compromise, configuration replacement, or code execution. This issue is fixed in version 1.1.2.

Updated: 2026-08-11

FreeCAD is a free and open-source multiplatform 3D parametric modeler. Prior to 1.1.2, the FEM Displacement Constraint task dialog in src/Mod/Fem/Gui/TaskFemConstraintDisplacement.cpp passes the xDisplacementFormula, yDisplacementFormula, and zDisplacementFormula fields of a Fem::ConstraintDisplacement object through TaskDlgFemConstraintDisplacement::accept() into Gui::Command::doCommand. The escaping helper neutralizes quotation marks but not backslashes, allowing crafted formula text to terminate the generated Python string and execute arbitrary Python code with the FreeCAD process's privileges when a victim accepts the dialog. This issue is fixed in version 1.1.2.

Updated: 2026-08-11

ffuf is a fast web fuzzer written in Go. Prior to 2.2.0, ffuf allows a malicious target server to cause an out-of-memory denial of service because the response size guard in pkg/runner/simple.go checks only the compressed Content-Length while io.ReadAll reads gzip, brotli, deflate, transparently decompressed, or chunked response bodies without a decompressed-size bound. This issue is fixed in version 2.2.0.

Updated: 2026-08-11

Faker generates massive amounts of fake data in the browser and Node.js. Prior to 10.5.0, the faker.helpers.fake method in src/modules/helpers/eval.ts allows attacker-controlled fake templates to access the Function constructor through fakeEval.resolveProperty when a function returns another function, enabling arbitrary JavaScript code execution. This issue is fixed in version 10.5.0.

Updated: 2026-08-11

Ente provides end-to-end encrypted cloud services and security tools. Prior to 2026.07.28, Ente 2of3 card format version 1 stored the secret byte length and 32-bit FNV-1a checksum in cleartext on every card, allowing someone with one card to test candidate secrets offline and recover low-entropy or predictable secrets. This issue is fixed in version 2026.07.28.

Updated: 2026-08-11

Django REST framework is a powerful and flexible toolkit for building Web APIs. Prior to 3.17.2, Django REST Framework's rest_framework/renderers.py AdminRenderer.render() uses override_method() to simulate GET and directly invokes view.get() without view.check_permissions() while rendering an invalid write request, allowing a 400 Bad Request HTML response to disclose data from a GET representation that the requester is not permitted to access. This issue is fixed in version 3.17.2.

Updated: 2026-08-11

Django REST framework is a toolkit for building Web APIs. Prior to 3.17.2, Django REST Framework's request.data parsing in rest_framework/request.py Request._parse() passes the underlying HttpRequest stream to JSONParser and FormParser for application/json and application/x-www-form-urlencoded bodies, bypassing Django's DATA_UPLOAD_MAX_MEMORY_SIZE protection and allowing oversized request bodies to consume additional memory and CPU. This issue is fixed in version 3.17.2.

Updated: 2026-08-11

electerm is an open-sourced terminal/ssh/sftp/telnet/serialport/RDP/VNC/Spice/ftp client. Prior to 3.15.120, electerm allows a malicious RDP server to write attacker-controlled content outside the selected save directory because the RDP clipboard download path in src/client/components/rdp/file-transfer.js passes the server-controlled CLIPRDR filename fileInfo.name to osResolve without sanitization. This issue is fixed in version 3.15.120.

Updated: 2026-08-11

electerm is an open-sourced terminal/ssh/sftp/telnet/serialport/RDP/VNC/Spice/ftp client. Prior to 3.15.186, electerm allows an authenticated WebSocket client to invoke unintended internal functions through client-controlled func values in upgrade-func in src/app/server/dispatch-center.js and handleFs in src/app/server/fs.js, exposing Upgrade and fsExport methods that can execute commands, open files, mutate the filesystem, or terminate the process. This issue is fixed in version 3.15.186.

Updated: 2026-08-11

electerm is an open-sourced terminal/ssh/sftp/telnet/serialport/RDP/VNC/Spice/ftp client. Prior to 3.15.120, electerm allows a malicious FTP or SFTP server to write attacker-controlled content outside the selected download directory because recursive transfers in src/client/components/file-transfer/transfer.jsx pass server-supplied file.name and folder.name values to resolve without sanitization. This issue is fixed in version 3.15.120.

Updated: 2026-08-11

electerm is an open-sourced terminal/ssh/sftp/telnet/serialport/RDP/VNC/Spice/ftp client. Prior to 3.15.120, electerm allows a malicious FTP or SFTP server to execute arbitrary commands when a user downloads a crafted folder and invokes Properties and Calculate Size because calcLocal in src/client/components/sftp/file-info-modal.jsx inserts the server-controlled folder name into a du -sh shell command without safely escaping single quotes. This issue is fixed in version 3.15.120.

Updated: 2026-08-11

electerm is an open-sourced terminal/ssh/sftp/telnet/serialport/RDP/VNC/Spice/ftp client. Prior to 3.15.120, electerm allows a malicious SFTP server to write attacker-controlled content outside the temporary directory because the server-controlled filename name used by editWithSystemEditor in src/client/components/sftp/file-item.jsx is interpolated into path.resolve without sanitization. This issue is fixed in version 3.15.120.

Updated: 2026-08-11

Claude Code Templates is a CLI tool for configuring and monitoring Claude Code. Prior to 1.29.4, the Claude Code Studio server launched by the --studio option in cli-tool/src/sandbox-server.js binds to all interfaces on port 3444, permits cross-origin requests, and requires no authentication. The POST /api/execute endpoint passes the prompt request-body field to executeLocalTask(), and POST /api/install-agent passes the agentName request-body field to a child process. The same unsafe agent field path is reachable from /api/execute through checkAndInstallAgent(). These attacker-controlled values reach child_process.spawn() with shell execution enabled, causing Node.js to construct a shell command in which metacharacters are interpreted. An attacker who can reach the port directly, or who convinces a developer running Studio to visit a malicious website, can execute arbitrary operating-system commands with the developer's privileges and compromise source code, credentials, and local data. This issue is fixed in version 1.29.4.

Updated: 2026-08-11

CVAT is an open source interactive video and image annotation tool for computer vision. From 2.17.0 until 2.72.0, a user with the Worker role can use predictable task-based request IDs with the lambda request retrieve and destroy endpoints to view automatic annotation requests for tasks or jobs the user cannot access and cancel requests initiated by other users. This issue is fixed in version 2.72.0.

Updated: 2026-08-11

CVAT is an open source interactive video and image annotation tool for computer vision. From 2.17.0 until 2.72.0, a user with write access to a CVAT job can submit a batch automatic annotation request to RequestViewSet.create with inconsistent task and job IDs, and because the task ID determines the single active request slot, block automatic annotation for another task whose ID is known. This issue is fixed in version 2.72.0.

Updated: 2026-08-11

Cursor is a code editor built for programming with AI. Prior to 3.0.0, Cursor IDE for macOS allows an agent running in Auto-Run Sandbox mode, when Docker Desktop and the Dev Containers CLI are installed, to launch a privileged container and mount Docker's virtiofs0, granting read and write access to the user's home directory and enabling host command execution with the user's privileges without an additional permission prompt. This issue is fixed in version 3.0.0.

Updated: 2026-08-11

Cursor is a code editor built for programming with AI. Prior to 3.1.2, Cursor IDE for macOS allows an agent running in Auto-Run Sandbox mode to replace a virtual environment's Python executable with a malicious wrapper that the Microsoft Python extension invokes outside the sandbox, allowing arbitrary host commands with the user's privileges, including modifying files outside the workspace and launching applications. This issue is fixed in version 3.1.2.

Updated: 2026-08-11

Coturn is a free open source implementation of TURN and STUN Server. Prior to 4.17.0, shutdown_client_connection() in src/server/ns_turn_server.c prematurely calls dec_quota() and releases bandwidth accounting during the first-stage close of a mobility-enabled allocation while preserving the allocation, relay socket, session, and mobility ticket, allowing an authenticated client to bypass --user-quota and --total-quota and exhaust relay ports. This issue is fixed in version 4.17.0.

Updated: 2026-08-11

Coturn is a free open source implementation of TURN and STUN Server. Prior to 4.17.0, turnports_allocate_even() in src/apps/relay/turn_ports.c marks the unused odd sibling port as TPS_TAKEN_ODD for an EVEN-PORT Allocate request with reservation bit R=0 even though no RTCP socket will release it, allowing an authenticated client to permanently exhaust the relay port pool and cause subsequent allocations to fail with STUN error 508. This issue is fixed in version 4.17.0.

Updated: 2026-08-11

Coturn is a free open source implementation of TURN and STUN Server. Prior to 4.16.0, dtls_server_input_handler() and create_new_connected_udp_socket() in src/apps/relay/dtls_listener.c retain OpenSSL dtls1_reassemble_fragment() state for a 35-byte fragmented ClientHello declaring a 650,000-byte handshake before cookie validation, allowing an unauthenticated remote sender using fresh UDP tuples to exhaust memory without TURN credentials, a completed handshake, a valid cookie, or source spoofing. This issue is fixed in version 4.16.0.

Updated: 2026-08-11

Coturn is a free open source implementation of TURN and STUN Server. Prior to 4.16.0, addr_less_eq() in src/client/ns_turn_ioaddr.c uses a component-wise comparison for native IPv6 min-max intervals in ioa_addr_in_range(), allowing an authenticated TURN client to relay to an IPv6 peer that is numerically within a configured non-prefix-aligned denied-peer-ip range but is classified as outside it. This issue is fixed in version 4.16.0.

Updated: 2026-08-11

Coturn is a free open source implementation of TURN and STUN Server. Prior to 4.13.1, good_peer_addr() in src/server/ns_turn_server.c uses ioa_addr_in_range() in src/client/ns_turn_ioaddr.c without canonicalizing IPv4-compatible, 6to4, and 64:ff9b::/96 NAT64 address forms, allowing an authenticated RFC 6062 TCP CONNECT relay client to bypass an IPv4 denied-peer-ip range when the Coturn host has a useful translation route. This issue is fixed in version 4.13.1.

Updated: 2026-08-11

PeerTube is an ActivityPub-federated video streaming platform. Prior to 8.1.6, ActorFollowModel.updateScore() interpolates the attacker-controlled ActivityPub actor inboxUrl into an SQL query, allowing an unauthenticated remote server to read and write PeerTube database tables, including oAuthToken.accessToken, and take over administrator accounts. This issue is fixed in version 8.1.6.

Updated: 2026-08-11

A Server-Side Request Forgery (SSRF) vulnerability existed in Lookyloo's PlaywrightCapture when the only_global_lookup option was enabled. PlaywrightCapture implements this option to prevent captures from accessing local, loopback, or otherwise non-public network resources. However, favicon retrieval was performed separately from the browser request-routing protections. Favicon URLs extracted from rendered HTML were resolved and subsequently fetched directly using an aiohttp.ClientSession. An attacker able to supply or control a web page processed by PlaywrightCapture could include a crafted favicon reference, for example pointing to a loopback address, private IP address, or another resource reachable only from the PlaywrightCapture host. When the page was processed, the favicon retrieval routine could issue an HTTP request to this destination despite only_global_lookup being enabled. This bypass could therefore be used to make the PlaywrightCapture host interact with internal network services that should not be reachable through a capture. Depending on the targeted service and its response, this could enable internal service discovery, access to internal resources, or interaction with HTTP endpoints available only from the capture infrastructure. The patch introduces a common URL validation routine and applies it to favicon retrieval. Direct non-global IP addresses, localhost, .local domains, malformed URLs, and other explicitly non-public destinations are rejected before the favicon request is performed. This fix is a complementary fix to CVE-2026-44439 - GCVE-0-2026-44439 - GHSA-687H-XW6F-Q2QW

Updated: 2026-08-11

Affected versions of MISP cti-transmute expose several state-changing account operations as GET requests: * /account/follow * /account/delete_notification * /account/mark_notification_read * /account/mark_all_read These endpoints require authentication, but before the fix they could be invoked with simple GET requests. That makes them susceptible to cross-site request forgery because a third-party site can induce the victim’s browser to send authenticated GET requests automatically. The patch converts the actions to POST or DELETE and updates the frontend to include an X-CSRFToken header, providing explicit CSRF protection for those state-changing operations.

Updated: 2026-08-11

Affected versions of cti-transmute improperly handle conversion-table values passed through the search highlighting feature. The highlight() function previously returned the underlying text directly when no search query was supplied, or performed a regex replacement that inserted <mark> tags without first escaping the original content. Because the resulting value is used by an HTML-rendering sink, malicious markup contained in conversion data could be interpreted as HTML rather than displayed as text. The fix introduces a shared highlightMatches() helper that first converts special characters such as <, >, &, and quotes into HTML entities. Only after escaping does the code insert the application-controlled <mark> element used for search highlighting.

Updated: 2026-08-11

Affected versions of cti-transmute contain an SSRF vulnerability in the /fetch_misp_event and /misp_search_events endpoints. The URL validation routine checked whether a supplied hostname was itself an IP literal and rejected private, loopback, link-local, or reserved IPs. However, ordinary domain names were accepted without resolving them first. An attacker could therefore use a hostname whose DNS record pointed to an internal address and cause the cti-transmute server to issue requests into its internal network. The commit explicitly states that anonymous callers could make the server request the internal target and read the response. The fix resolves hostnames using socket.getaddrinfo(), checks that every resolved address is globally routable, and additionally places @login_required on both affected MISP fetch/search routes.

Updated: 2026-08-11

Affected versions of cti-transmute allow a tag's icon value to be stored and later interpolated into HTML through Vue's v-html. The helper mapIcon() previously constructed an HTML string directly from the icon value: <i class="fas fa-${name}"></i> Because the icon is user-supplied, a crafted value could break out of the intended markup and inject attacker-controlled HTML. When the affected tag was later rendered, including on the administrative triage interface, the payload could execute in the viewer's browser. The patch mitigates the issue at multiple layers: v-html is replaced with Vue :class binding, mapIcon() now returns only a constrained FontAwesome class string, and the backend validates icons against the FontAwesome catalogue or a strict [a-z0-9-]{1,40} slug pattern before storing them.

Updated: 2026-08-11

Affected versions of cti-transmute insufficiently validate saved graph configuration data. Graph configurations can contain style properties that are later consumed by Pivotick, and Pivotick interprets svgIcon as HTML. Because saved configurations may be created by one user and later displayed to other users—including administrators—a malicious user could store a crafted svgIcon value that executes script in another viewer’s browser. The fix introduces a strict configuration schema on both the server and client. Only known properties are accepted, style entries are limited to shape, color, and size, and dangerous properties such as svgIcon and iconClass are explicitly rejected. Existing stored configurations are also sanitized when listed and again before being applied in the browser.

Updated: 2026-08-11

Affected versions of cti-transmute render data obtained from a remote MISP instance into the event-browser interface using HTML interpolation. Because fields such as event IDs, event information, organization names, tags, tag colors, TLP labels, distribution labels, and error/flash text may be controlled by the remote MISP server, a malicious or compromised remote instance could return crafted values that inject HTML or script-capable content into the cti-transmute interface. The patch explicitly notes that remote-derived values must not reach innerHTML, and replaces string-built rows and badges with DOM nodes populated through textContent. It also restricts remote-controlled tag colors to six-digit hexadecimal values, preventing malicious CSS values such as url(...).

Updated: 2026-08-11

Affected versions of cti-transmute fail to HTML-escape attacker-controlled values used in ECharts Sunburst and Treemap tooltip formatters. Slice names may originate directly from STIX or MISP data, including STIX types, relationship_type, pattern prefixes, and MISP category/type values. Since ECharts interprets the formatter return value as HTML, crafted values can inject markup or script-capable content into the tooltip. An attacker who can cause malicious conversion data to be processed can therefore inject content that executes when another user views the visualization and hovers over the affected slice. The patch replaces direct interpolation with dedicated formatter functions that call escapeHtml() on p.name, p.data.value, and p.value.

Updated: 2026-08-11

Affected versions of cti-transmute allow authenticated users to add or remove emoji reactions on comments without first checking whether those users are authorized to view the target comment. The vulnerable react() handler passed an attacker-controlled comment_id directly to comments_repo.toggle_reaction() after only validating that the ID existed syntactically and that the requested emoji was permitted. Because comment-level visibility was not enforced, a user who could identify the ID of a private or otherwise inaccessible comment could modify reaction state on that comment despite lacking permission to access it. The fix retrieves the target comment, rejects missing or deleted comments, retrieves its associated conversion, and enforces access.can_see_comment(current_user, comment, conversion). Unauthorized requests now receive HTTP 403.

Updated: 2026-08-11

Affected versions of cti-transmute fail to apply comment-level access-control rules when generating evaluation report exports. Although normal comment retrieval filters comments according to conversion visibility, comment privacy, ownership, authorship, and administrative privileges, build_evaluation_report() previously included all evaluation comments without applying those rules. Consequently, a user who was authorized to view a conversion could export its evaluation report as Markdown or PDF and obtain private evaluation comments that should only have been visible to the conversion owner, the comment author, or an administrator. The leaked report data also contained the comment author's name. The fix passes the requesting user into the report builder and filters every evaluation comment using the shared access.can_see_comment() authorization function.

Updated: 2026-08-11