Just published.
The newest CVEs, straight from CVE.org and NVD.
-
CVE-2026-119205
pypdf is a free and open-source pure-python PDF library. Prior to 6.21.4, pypdf can enter an infinite loop when cloning or duplicating a crafted cyclic DictionaryObject. Clone traversal checked generated destination references rather than the original source references, allowing some source cycles to evade the visited-object check. This issue is fixed in version 6.21.4.
-
CVE-2026-119206
pypdf is a free and open-source pure-python PDF library. Prior to 6.21.3, pypdf can take excessive time to retrieve page labels from a crafted PDF containing a flat page-label number tree with many ranges. The page_labels property resolves each page independently and repeatedly scans the same /Nums array from the beginning, causing work to grow with both the page count and the number of label ranges. This issue is fixed in version 6.21.3.
-
CVE-2026-119207
pypdf is a free and open-source pure-python PDF library. Prior to 6.21.3, pypdf can consume excessive memory when strict mode loads a crafted invalid PDF dictionary. Nested parsing failures add repeated exception text to newly raised errors, allowing a small malformed dictionary to amplify into very large error messages. This issue is fixed in version 6.21.3.
-
CVE-2026-119194
pyasn1 is a generic ASN.1 library for Python. From 0.3.3 until 0.6.5, schemaless container decoding stores every decoded component's TagSet in a growing set even though only one-versus-multiple tag types must be distinguished. Attacker-controlled BER, CER, or DER input contains many distinct long tag IDs whose TagSet hashes collide. Each later set insertion compares against an increasing number of earlier colliding tags in ConstructedPayloadDecoderBase._decodeComponentsSchemaless, producing quadratic CPU work while decoded values and order remain valid. Repeated inputs can exhaust decoder CPU and reduce service availability. This issue is fixed in version 0.6.5.
-
CVE-2026-119195
pyasn1 is a generic ASN.1 library for Python. Prior to 0.6.5, constructed string and indefinite-length ANY decoding repeatedly concatenates each fragment onto a growing bytes value. Attacker-controlled BER or CER data can use many constructed OCTET STRING or character-string fragments. DER rejects constructed OCTET STRING without a schema, but accepts its definite-length form when asn1Spec selects univ.OctetString; it also accepts constructed character strings such as UTF8String without a schema, even though those encodings are not valid DER. The separate indefinite-length ANY path is reachable in BER and CER when asn1Spec selects univ.Any; DER rejects indefinite lengths. OctetStringPayloadDecoder uses growing header concatenation and AnyPayloadDecoder.indefLenValueDecoder uses growing chunk concatenation, repeatedly copying an increasing prefix and causing quadratic work. A sufficiently fragmented value can occupy decoding workers and cause CPU exhaustion without confidentiality or integrity impact. This issue is fixed in version 0.6.5.
-
CVE-2026-119196
pyasn1 is a generic ASN.1 library for Python. Prior to 0.6.5, constructed BIT STRING decoding repeatedly shifts and copies the complete bit value accumulated from earlier fragments. Attacker-controlled BER or CER input contains many constructed BIT STRING fragments; schema-driven DER also accepts a definite-length constructed BIT STRING. BitStringPayloadDecoder repeatedly calls BitString.fromOctetString with the prior bitString as prepend, shifting an ever-growing SizedInteger for each fragment and causing quadratic work. Large fragmented values can exhaust CPU during decode() without requiring later conversion or formatting. This issue is fixed in version 0.6.5.
-
CVE-2026-119197
pyasn1 is a generic ASN.1 library for Python. Prior to 0.6.5, binary REAL decoding builds a large mantissa one octet at a time by shifting a growing integer and slicing the remaining byte string. Attacker-controlled ASN.1 includes a binary-encoded REAL with a large mantissa and reaches BER, CER, or DER decode(), with or without a schema containing REAL. RealPayloadDecoder.valueDecoder repeatedly performs a left shift, OR, and suffix copy for every mantissa octet, making decode time quadratic in mantissa length. A comparatively small REAL value can hold a decoding worker on one CPU core for seconds or longer and cause availability loss. This issue is fixed in version 0.6.5.
-
CVE-2026-9536
A heap-based buffer overflow vulnerability exists in the vtkWindBladeReader FindVariableOffsets functionality of vtk (version(s): 9.6.1). A specially crafted .wind can lead to a buffer overflow. An attacker can provide a malicious file to trigger this vulnerability.
-
CVE-2026-9535
A out-of-bounds read vulnerability exists in the vtkChacoReader InputGraph2 functionality of vtk (version(s): 9.6.1). A specially crafted chaco can lead to an out-of-bounds read. An attacker can provide a malicious file to trigger this vulnerability.
-
CVE-2026-118779
Forge (also called node-forge) is a native implementation of Transport Layer Security in JavaScript. Prior to 1.4.1, the affected implementation used pem.decode(), lib/pem.js, missing END line to process the relevant input or security decision. node-forge used polynomial-time regular expressions in pem.decode(), including when searching for a matching END line and parsing encapsulated headers. An unauthenticated attacker able to supply PEM input to any high-level parser could consume excessive CPU with a long malformed message and cause denial of service. This issue is fixed in version 1.4.1.
-
CVE-2026-51851
modelscope agent-zero 1.7, 1.8, 1.9, and 1.10 is vulnerable to Directory Traversal in python/helpers/file_browser.py:FileBrowser.save_text_file. The save_text_file method accepts user-controlled file paths without normalization or validation, allowing path traversal attacks.
-
CVE-2026-118769
Botan is a C++ cryptography library. From 2.1.0 until 3.13.0, the Botan C API function botan_block_cipher_block_size converted a size_t block size to int without a bounds check. In an application that accepts external algorithm parameter strings and uses the FFI raw ECB mode, an attacker can select the variable-block-size Lion cipher so that the reported size is truncated, the caller allocates insufficient memory, and later block processing causes memory corruption. The C++ API and normally used C API modes such as CBC and CTR are unaffected. This issue is fixed in version 3.13.0.
-
CVE-2026-118766
Botan is a C++ cryptography library. Prior to 3.13.0, Botan's online OCSP client allowed a single HTTP redirect while posting an OCSP request to the URI in a certificate's Authority Information Access extension. A malicious OCSP responder or an on-path attacker could redirect the client to make one blind request to an otherwise unreachable internal HTTP service. The primary impact is blind SSRF that might affect internal services with non-idempotent GET handlers, and response timing can provide a coarse internal-service oracle. This issue is fixed in version 3.13.0.
-
CVE-2026-118767
Botan is a C++ cryptography library. From 1.11.22 until 3.13.0, Botan's Python binding converted Python strings containing U+0000 to const char* values without consistently rejecting the embedded NUL, causing bcrypt and check_bcrypt to process only the prefix before the NUL. Applications could validate a longer password but hash a shorter effective value, allowing application password-policy bypass; scrypt and argon2 are unaffected. This issue is fixed in version 3.13.0.
-
CVE-2026-118768
Botan is a C++ cryptography library. From 1.11.34 until 3.13.0, the botan ocsp_check command requested an OCSP response from a certificate's Authority Information Access URL without verifying the response signature or authorizing the signer. A malicious responder or on-path attacker could return a matching and timely GOOD response with an empty, invalid, or unauthorized signature, causing a revoked certificate to produce exit status 0 and OCSP check OK output. Only the command-line interface is affected. This issue is fixed in version 3.13.0.
-
CVE-2026-62374
SD JWT JS is a JavaScript implementation of the Selective Disclosure JWT (SD-JWT) specification. In version 0.19.0 and earlier, `@sd-jwt/decode`, also used by `@sd-jwt/core`, silently drops a tampered Disclosure whose digest no longer matches the Issuer-signed `_sd` array and returns a reduced claim set, reporting success while the Issuer signature remains valid. Both a tampered Disclosure that remains valid JSON and a Disclosure containing invalid UTF-8 bytes reach this silent-drop behavior. In the latter case, the `TextDecoder / Buffer` decoding path substitutes U+FFFD replacement characters, recovering syntactically valid disclosure data instead of rejecting the malformed input. This violates RFC 9901 section 7.1 step 5, which requires rejection of unreferenced Disclosures, and section 7.1 step 3 / section 4.2.1, which require rejection of invalid Disclosures. A malicious holder can exploit this omission primitive, including through `SDJwtInstance.verify()` when key binding is required. An on-path attacker can also tamper with a Disclosure when the relying party does not enforce key binding; enforced key binding rejects on-path tampering because it invalidates the KB-JWT `sd_hash` binding. If the relying party defaults to allowing access when a constraining age, role, or restriction claim is absent, the silent claim reduction can change an authorization decision without an error. This issue is fixed in `@sd-jwt/core` version 0.20.0; no patched version of the standalone `@sd-jwt/decode` package is identified.
-
CVE-2026-118763
sigstore-js provides JavaScript libraries for interacting with Sigstore services. Prior to 0.7.2, packages/oci/src/registry.ts contains a RegistryClient.signIn() method that stores registry Authorization and Docker configuration headers as defaults on a shared HTTP client, while RegistryClient.uploadBlob() sends a fresh PUT to an absolute blob-upload Location supplied by the registry without checking that its origin or scheme matches the registry. A malicious or compromised registry can therefore direct a user pushing an artifact through attachArtifactToImage() to an attacker-controlled host and receive the registry bearer token, Basic credentials, or custom authentication headers. Redirect-time header stripping does not prevent the leak because the cross-host PUT is a new explicit request rather than a followed redirect. This is a residual of CVE-2026-59891, whose fix changed credential selection in credentials.ts but did not constrain where credentials are sent from registry.ts. Version 0.7.2 fixes the issue by separating registry credentials from client-wide defaults and sending them only to the registry origin.
-
CVE-2026-118764
minimatch is a minimal matching utility for converting glob expressions into JavaScript RegExp objects. Starting in 7.0.0 and prior to 11.0.1, src/index.ts implements firstPhasePreProcess() so that, at optimizationLevel 2 or higher, it repeatedly expands attacker-controlled **/.. pattern segments inside a fixpoint loop, allowing generated globParts entries to be processed and expanded again. secondPhasePreProcess() then compares the expanded entries pairwise, so a short malicious glob can cause excessive CPU and memory consumption during pattern compilation. This issue is fixed in version 11.0.1.
-
CVE-2026-118765
Botan is a C++ cryptography library. From 3.7.0 until 3.13.0, an application that calls AutoSeeded_RNG::clear() and then calls randomize with an empty output buffer while the generator is unseeded can cause AutoSeeded_RNG to mark itself seeded without consulting the configured entropy source. On builds without BOTAN_HAS_SYSTEM_RNG, the fallback uses only a clock value and process ID, so a later non-empty request can produce output from predictable state. This issue is fixed in version 3.13.0.
-
CVE-2026-107057
An issue in AndBible versionCode 1111 allows attackers to execute arbitrary code via uploading a crafted module zip.
-
CVE-2026-107052
Insecure permissions in the AIDL interface of ProjectMAXS MAXS Module FileRead versionCode 27 allows unauthorized attackers to arbitrarily reead files outside of the sandbox via providing an absolute path.
-
CVE-2026-93570
Szafir and Szafir Weryfikująca are vulnerable to XML External Entity (XXE) injection. When a user signs or verify a crafted XML file containing a `DOCTYPE` with an external DTD reference, the parser processes external entities, allowing an attacker to read arbitrary files from the victim's filesystem. This issue was fixed in Szafir version 2.0 build 708. Szafir Weryfikująca is no longer supported (End of Life) and has not been fixed.
-
CVE-2026-119131
Wasmtime 46.0.0 before 48.0.6 and 49.0.0 before 49.0.3 contains a denial of service vulnerability in wasmtime-wasi-http that lets guests crash the host by sending long request URLs. Guests can use set-path-with-query with a '?' after byte 65535, overflowing a u16 index in the http crate and triggering a host panic.
-
CVE-2026-119133
Wasmtime 47.0.0 before 48.0.6 and 49.0.0 before 49.0.3 contains a memory misalignment flaw in VMContext layout that causes undefined behavior when modules contain an odd number of exception tags. A WebAssembly module with an odd tag count misaligns subsequent host fields in embedders with exceptions enabled, although no exploitation path is currently known.
-
CVE-2026-111731
P-Net before 1.0.3 enters an infinite loop when handling a read request for the PDPortDataCheck record with subslot 0 (whole module). The loop over the ports never advanced to the next port, so the request never completes and the device stops responding. A remote unauthenticated attacker can use this to cause a denial of service.
-
CVE-2026-119051
flat through 6.0.1 contains an uncontrolled resource allocation vulnerability in unflatten() that allows attackers to create huge sparse arrays via unbounded numeric keys. Attackers can supply flattened keys such as a.4294967294, causing iteration with forEach, spread or JSON.stringify to block the Node.js event loop.
-
CVE-2026-119050
Agno through 3.1.11 contains an unsafe reflection vulnerability in PandasTools that allows attackers to invoke arbitrary pandas functions by supplying model-controlled function names to getattr(). Attackers can use prompt injection to make the agent call pandas.read_pickle on attacker-controlled paths for code execution, or call other readers to read local files.
-
CVE-2019-25781
Python-Markdown 3.0 through 3.11.2 contains an algorithmic complexity vulnerability in inline link parsing that allows remote attackers to exhaust CPU through unmatched brackets. Attackers can submit short Markdown containing repeated [ or ![ characters, forcing LinkInlineProcessor.getText() and getLink() into quadratic scans that keep workers busy for minutes.
-
CVE-2026-119027
oapi-codegen 2.5.0 before 2.8.1 contains a code injection vulnerability that allows attackers to inject Go code by supplying a crafted servers[].x-go-name value emitted unescaped by the server-urls template. Attackers who control an OpenAPI document can define an injected func init() that executes when the victim builds and imports the generated package.
-
CVE-2026-119028
oapi-codegen 2.0.0 before 2.8.1 contains a code injection vulnerability that allows attackers to inject Go code by supplying non-literal enum entries in integer, number, or boolean schemas. Attackers controlling an OpenAPI document can craft an enum entry that closes the const block in constants.tmpl and injects a func init(), executing code when the generated package is built and imported.
-
CVE-2026-118905
Node Version Manager (nvm) is a POSIX-compliant shell function for managing multiple node.js versions. In versions 0.31.2 through 0.40.5, `nvm_ls` filters installed versions with a sed program that embeds the requested version pattern in a `\#...#` address, escaping `.` (and, since 0.40.4, `#`) but not backslash. A version string containing `#` (or `\#`) therefore closes the address early, and the rest of it is parsed as sed commands: with GNU sed, the `e` command runs a shell command as the user; with BSD sed, the `w` and `r` commands write and read files. This is reachable whenever an attacker-influenced version string is passed to `nvm use`, `nvm which`, `nvm ls`, `nvm version`, `nvm run`, or `nvm exec`, or to `nvm-exec` via `NODE_VERSION`, and at least one version is installed. nvm's own `.nvmrc` parser strips comments, so a bare `nvm use` is not affected; however, the bash `cdnvm` and zsh `load-nvmrc` auto-switch hooks documented in the README read `.nvmrc` as-is and pass it to `nvm ls` and `nvm version`, so a user with one of those hooks installed runs the attacker's command just by changing into a directory under a malicious `.nvmrc`; scripts that run `nvm use "$(cat .nvmrc)"` are affected the same way. Version 0.40.6 passes any version pattern containing `#` or a newline through the `.nvmrc` content parser first, so no `#` reaches the sed program.
-
CVE-2026-66089
In openEQUELLA before 2026.1, the SOAP API parses XML supplied in the stopEdit operation with a parser that resolves external entities. This allows an attacker to read arbitrary files accessible to the openEQUELLA service or cause the server to issue requests to attacker-specified locations (SSRF).
-
CVE-2026-118590
A security vulnerability has been detected in nipy nibabel up to 5.4.2. The affected element is the function array_from_file of the file nibabel/volumeutils.py of the component Header Handler. The manipulation leads to uncontrolled memory allocation. The attack is possible to be carried out remotely. The pull request to fix this issue awaits acceptance.
-
CVE-2026-118694
python-markdown2 through 2.5.5 contains a denial of service vulnerability that allows remote users to exhaust CPU by submitting Markdown with unclosed inline link destinations. Attackers can submit repeated '[a](' or 's handler class and args fields. Attackers can publish a malicious bundle whose logging.conf runs arbitrary Python at workflow construction when a victim loads or runs the bundle.
-
CVE-2026-118684
MONAI before 1.6.1 contains an insufficient integrity verification vulnerability in check_hash() that silently returns True when hash_val is None, with MD5 as the default hash type. Attackers performing man-in-the-middle or registry compromise can substitute bundles, datasets, or model archives downloaded without hashes, enabling malicious file delivery and potential code execution.
-
CVE-2026-118259
opentelemetry-java is the Java implementation of the OpenTelemetry API for recording telemetry, and SDK for managing telemetry recorded by the API. Prior to 1.67.0, W3CTraceContextPropagator.extract() uses regex delimiter processing that can exhibit quadratic backtracking while parsing tracestate headers containing long runs of spaces or tabs without a following comma. An unauthenticated attacker whose headers reach extraction can repeatedly block request or event-loop threads and cause excessive CPU consumption, with practical impact bounded by ingress header-size limits, proxy configuration, and threading. This issue is fixed in version 1.67.0.
-
CVE-2026-116794
russh is a Rust SSH client and server library that includes an SSH agent protocol implementation. From 0.50.0-beta.12 until 0.65.0, the agent server's Connection::try_sign path in russh/src/keys/agent/server.rs recognizes SSH_AGENT_CONSTRAIN_CONFIRM but returns a hard-coded successful result instead of calling Agent::confirm. In applications that override Agent::confirm to require user approval or another policy check, a process that can reach the agent socket, including the remote side of a forwarded agent, can use a key added through ssh-add -c without that check and can authenticate or sign data as the user while the key remains loaded. The separate Agent::confirm_request hook still runs for every signing request, so checks implemented there remain effective, and implementations that retain the default always-successful Agent::confirm do not lose an approval check through this regression. This issue is fixed in version 0.65.0.
-
CVE-2026-117500
HTTP::Body versions before 1.24_01 for Perl allow CPU exhaustion via regex backtracking on the filename of a multipart upload. For every file part, the multipart parser finds the suffix of the uploaded filename by matching the default $basename_regexp, qr/[^.]+(\.[^\\\/]+)$/, against the whole basename. A basename made of many dots and ending in a backslash makes the regex backtrack in time quadratic in its length. A remote attacker can exhaust CPU with one request, as a basename of about 32 KB takes several seconds to match and every file part is matched separately.
-
CVE-2026-117501
HTTP::Body versions before 1.24_01 for Perl allow memory exhaustion via unbounded buffering of chunked request bodies. In chunked mode, used when no content length is given, add() holds each chunk in memory until all of it has arrived. A line that is not a chunk-size line, a size line without CRLF, or a size too large to fit an integer leaves the parser waiting forever, and every later add() grows the buffer. A remote attacker can exhaust memory in applications that pass chunked request bodies to HTTP::Body.
-
CVE-2026-116267
MIME::Parser versions before 5.521 for Perl allow memory exhaustion via indented error messages from a truncated, deeply nested message. Each stored parser diagnostic is indented by three spaces per nesting level. When a message nested N levels deep ends without its closing boundaries, every level logs an error at its own depth, so the stored diagnostics grow with the square of N. The defaults place no limit on nesting depth.
-
CVE-2026-116270
MIME::Words versions before 5.521 for Perl allow memory exhaustion via repeated "=?" sequences in decode_mimewords. For each "=?" that does not start a complete encoded word, decode_mimewords appends an error message containing the whole input string to $@. A boundary of 8,000 "=?" pairs (16,000 bytes) adds 8,000 copies of itself to $@, about 128 MB. From version 5.518, MIME::Parser decodes every multipart boundary this way, so a remote sender can exhaust the memory of the parsing process with one Content-Type header.
-
CVE-2026-116271
MIME::Parser::FileUnder versions before 5.521 for Perl reuse a pre-created output directory and follow symlinks in it, letting local users redirect attachment writes. Each message is written to a directory named msg-<time>-<pid>-<n> under the output base, and an existing directory of that name is used as is. A file name counts as unused when nothing exists at that path, which holds for a dangling symlink, and the body file is then opened in a way that follows the link. When the output base is writable by other users, as with output_under("/tmp"), a local user can pre-create the directory with a symlink named like an attachment. The parser then writes the attachment to the link target with the privileges of the parsing process.
-
CVE-2026-116272
MIME::Field::ParamVal versions before 5.521 for Perl do not flag a duplicate boundary parameter whose name contains a nested comment. A parameter written boundary((c))="A" or boundary(\))="A" is stored under that literal name rather than as boundary. A second boundary parameter is then not reported as a duplicate, and MIME::Parser splits the message on it. Readers that skip comments as RFC 5322 allows take the first boundary instead. A remote sender can hide an attachment from a filter that relies on ambiguous_content() while the recipient's mail client shows it.
-
CVE-2026-116273
MIME::Parser versions before 5.521 for Perl set no limit on the number of message parts by default, allowing memory exhaustion. The max_parts option defaulted to -1, which disables the limit, so a parser created with default settings builds an entity for every part however many a message has. Version 5.521 sets max_parts to 1000 and max_depth to 64 by default. Setting either to -1 still removes the limit.
The newest published CVE records from CVE.org and NVD, refreshed every few minutes, with severity, exploit odds and fix status.