Intelligence
Explorer.

Godot Engine through 4.7.2 has an out-of-bounds read in the optimized translation loader (core/string/optimized_translation.cpp, OptimizedTranslation::get_message). Offsets read from the hash table of a crafted .translation resource are used as indices into the bucket table without bounds validation; only a single sentinel value is rejected. A crafted file can point an index far past the allocated buffer. Depending on the values, the result is a crash, or a string lookup (tr()) returns raw bytes from elsewhere in the process heap as the translated string, disclosing process memory. The defect is reached when the product or an exported game loads and performs lookups in a crafted translation file. NOTE: the Supplier's position is that the research has identified "no security issues" because the required attacker access is outside of the current Godot Engine threat model.

Godot Engine through 4.7.2 has a heap out-of-bounds write in the binary resource loader (core/io/resource_format_binary.cpp, ResourceLoaderBinary::get_unicode_string). A string length is read from the resource file into a signed 32-bit integer and compared against the destination buffer size with a signed comparison, which conditionally resizes the buffer. A negative length value skips the resize; the same value is then passed to the copy routine as an unsigned byte count, where it sign-extends to a huge value, copying file-controlled data far past the undersized heap allocation. The defect is reached when the product or an exported game loads a crafted binary resource file (.res). NOTE: the Supplier's position is that the research has identified "no security issues" because the required attacker access is outside of the current Godot Engine threat model.

Godot Engine through 4.7.2 has a heap out-of-bounds write in the Radiance HDR (.hdr) image decoder (modules/hdr/image_loader_hdr.cpp). The run-length decoding loop expands RLE instructions read from the file into a buffer sized from the image dimensions, without validating that an instruction's repeat/copy count fits the remaining capacity of the row buffer. A crafted .hdr file therefore writes attacker-supplied bytes past the end of the heap allocation, with the overflow length and content controlled by the file. The defect is reached automatically when the product imports a crafted .hdr (e.g., when opening a project containing one), and at runtime in exported projects that load untrusted .hdr images. NOTE: the Supplier's position is that the research has identified "no security issues" because the required attacker access is outside of the current Godot Engine threat model.

Portkey AI Gateway 1.14.0 through 1.15.2 contains a server-side request forgery vulnerability that allows unauthenticated attackers to reach loopback and Docker host services because isValidCustomHost trusts localhost-style hosts when TRUSTED_CUSTOM_HOSTS is unset. Attackers can send /v1 requests with a crafted x-portkey-custom-host header, such as via the ollama provider, to make the gateway call internal services and return their responses.

TurboDocx html-to-docx through 1.23.1 contains a server-side request forgery vulnerability that allows attackers supplying HTML to make the server fetch internal URLs via img src attributes. Attackers can reference loopback, private, or cloud metadata addresses that bypass the loose isValidUrl() check, and read responses embedded in the generated DOCX word/media folder.

Janus WebRTC Server through 1.4.3 contains a path traversal vulnerability in the Record&Play plugin that allows unauthenticated attackers to write files outside recordings_path via the record request filename. Attackers can supply filenames with path separators to create or truncate files ending in -audio.mjr, -video.mjr or -data.mjr anywhere the Janus process can write outside protected folders.

unshield through 1.6.2 contains a heap out-of-bounds read vulnerability in unshield_get_file_table() that allows attackers to read past heap allocations via unchecked cabinet header offsets. Attackers can supply a crafted InstallShield cabinet with out-of-range file_table_offset or large file counts to crash the unshield tool or library consumers.

MuJS 1.3.6 before 1.3.9 contains a heap out-of-bounds read and write vulnerability in Array.prototype.sort that allows attackers to corrupt memory by inflating a simple array's length. Attackers can supply JavaScript that enlarges a dense array's length and calls sort, causing Ap_sort_cmp and Ap_sort_swap to access slots past the buffer and crash the process.

libsmb2 through 6.2 contains an integer underflow vulnerability in decode_acl that allows malicious SMB servers to trigger heap out-of-bounds reads via crafted security descriptors. Attackers controlling a server or intercepting unsigned sessions can send an AclSize below 8 in a QUERY_INFO reply, bypassing ACE bounds checks and crashing the client.

libsmb2 through 6.2 contains a heap out-of-bounds read vulnerability in smb2_decode_file_fs_volume_info and smb2_decode_file_fs_attribute_info that allows malicious SMB servers to read past reply buffers. Attackers controlling a server or intercepting unsigned sessions can send crafted QUERY_INFO replies with missing or oversized name lengths to crash the client.

libsmb2 through 6.2 contains a heap out-of-bounds read vulnerability in smb2_decode_file_all_info that allows malicious SMB servers to crash clients by sending an unchecked FileNameLength. Attackers controlling the server or performing man-in-the-middle on unsigned sessions can return crafted FileAllInformation replies to smb2_stat or smb2_fstat calls, crashing the client.

CGNS through 4.5.2 contains a stack out-of-bounds read vulnerability in ADFI_read_node_header that allows attackers to crash applications by supplying crafted ADF node headers. Attackers can craft .cgns or .adf files whose non-NUL-terminated end tag causes ADFI_stridx_c to scan past disk_node_data, crashing tools like cgnslist.

vgmstream through r2117 contains a stack out-of-bounds read vulnerability in the G.722.1/Siren14 decoder in g7221_lib.c, whose bit reader refills without checking the data_u32 array end. Attackers can supply crafted .s14 or .sss files that drive decode_vector_quantized_mlt_indices() past the stack buffer during format detection, crashing the application.

ELFIO through 3.12 contains a NULL pointer dereference vulnerability in relocation_section_accessor get_entry() that allows attackers to crash applications via an unvalidated sh_link index. Attackers can supply a crafted ELF file whose SHT_REL or SHT_RELA section has a dangling sh_link, causing find_hash_section() to dereference null and deny service.

Sofia-SIP through 1.13.18 contains a heap out-of-bounds read vulnerability in msg_multipart_parse() that allows unauthenticated remote attackers to crash the process by sending oversized MIME boundaries. Attackers can send an INVITE with a multipart Content-Type whose boundary exceeds the body length, causing memcmp to read past the heap buffer and crash the process.

Khronos NNEF-Tools nnef through 1.0.10 and nnef-tools through 1.0.11 contain a path traversal vulnerability that allows attackers to write files outside the model directory via crafted variable labels. Attackers can supply a graph.nnef with '../' sequences or absolute paths in labels to create or overwrite '.dat' files when users convert, save, or generate weights.

miniupnpc through 2.3.3 contains a one-byte heap out-of-bounds read vulnerability in the minixml parseelt function's namespace-prefix handling that skips an end-of-buffer check after advancing past ':'. A malicious or spoofed UPnP gateway on the local network can send a device description or SOAP reply ending in ':' within an element name to crash the client.

miniupnpc through 2.3.3 contains a one-byte heap out-of-bounds read vulnerability in the minixml parseatt function that allows malicious UPnP gateways to read past buffer bounds via crafted replies. Attackers on the local network can spoof a gateway serving device descriptions or SOAP responses ending in '=' to crash the client.

libzvbi through 0.2.45 contains an out-of-bounds write vulnerability in the Teletext decoder that allows attackers to corrupt memory by supplying unvalidated DRCS pixel-mode values in X/28/3 packets. Attackers can send a crafted Teletext stream decoded via vbi_decode with every slot set to DRCS_MODE_6_5_4, causing convert_drcs to overflow chars[48][60] and crash the application.

3proxy from 0.9.7 through 1.0.1, fixed in commit 206c923, contains an uninitialized stack memory disclosure vulnerability that allows unauthenticated attackers to leak stack bytes via the SOCKS4 reply path when the external-NAT option (-Ne) is configured. Attackers can send a client probe that fails the SOCKS version check to trigger a reply containing uninitialized bytes from the sin structure, exposing high bytes of a userspace pointer that weakens ASLR protections.

mise manages dev tools like node, python, cmake, and terraform. Prior to 2026.7.14, in the default non-safe mode, mise loads .mise.toml [settings] from an untrusted local project before trust approval and strip_local_only_settings removes only fields marked global_only. The unix_default_file_shell_args, unix_default_inline_shell_args, windows_default_file_shell_args, and windows_default_inline_shell_args settings were not marked global_only, allowing a repository to replace the default interpreter arguments used by trusted tasks that have no explicit shell, postinstall directives, template exec operations, or hooks that fall back to the default shell. When mise appends the trusted command after the attacker-controlled interpreter arguments, the attacker-selected command can execute first with the developer account's privileges even though the local configuration was never trusted. The separate path-sanitization hardening note in the advisory is outside this vulnerability. This issue is fixed in version 2026.7.14.

Zulip is an open-source team collaboration tool. In Zulip Server versions through 12.1, GenericOpenIdConnectBackend matches the identity provider email claim to an active account's delivery_email without rejecting an explicitly false email_verified claim. If a configured identity provider returns an email address that the authenticating user does not control together with email_verified=false, Zulip can log the user into the existing account for that address. Exploitation requires a misconfigured trusted identity provider, and deployments whose authoritative identity provider only emits email addresses controlled by the authenticating user are not affected. This issue is fixed in version 12.2.

OmniFaces is a utility library for Faces. In OmniFaces versions before 1.14.3, 2.7.33, 3.14.23, 4.7.12, and 5.4.2, CombinedResourceHandler accepts a structurally valid deflated and base64 encoded combined resource ID without verifying that the server issued it. A forged ID can cause unbounded decoded output, add unique reconstructed IDs to a process-lifetime cache, or combine resources other than stylesheets and scripts. Because jakarta.faces.RESOURCE_EXCLUDES is not applied by this custom resolution path, a forged ID can disclose Facelets source from the webapp resource tree or META-INF/resources. In a deployment configured with the wildcard CDN mapping, a forged ID that selects omnifaces.graphic can also cause a blind, deployment-dependent server-side fetch whose destination derives partly from the request Host header. The flaw does not provide arbitrary file read or establish arbitrary-scheme or arbitrary-destination SSRF. The shared authenticity failure permits memory exhaustion and limited resource disclosure. This issue is fixed in versions 1.14.3, 2.7.33, 3.14.23, 4.7.12, and 5.4.2.

Angular is a development platform for building mobile and desktop web applications using TypeScript/JavaScript and other languages. From 1.20.0 until 1.21.1, the Angular DevTools background service worker accepts chrome.runtime.sendMessage runtime messages from untrusted web pages or third-party content scripts without adequately verifying the message source. An untrusted sender that obtains a valid randomized victim tabId can direct the extension to use chrome.debugger to attach to another browser tab and trigger Chrome DevTools Protocol breakpoints. This can display the debugging banner, pause JavaScript execution, and prevent interaction with the target tab, but it does not provide remote code execution, privilege escalation, or cross-origin access to memory or credentials. This issue is fixed in version 1.21.1.