CVE-2026-120757
Description détaillée
HDF5 through 2.2.0 contains a heap buffer over-read vulnerability in H5L__extern_traverse() in src/H5Lexternal.c, the default traversal callback for external links. The callback locates the target file name and object path within the link’s user data using strlen() without validating the buffer against its stored size (udata_size, which was marked unused). A crafted HDF5 file containing an external link whose data is not NULL-terminated, or is too short to hold both strings, causes strlen() to read past the end of the udata heap buffer, and the derived object path pointer may also point outside the buffer. The overrun is unbounded and continues until a zero byte is found. The flaw is triggered when the malicious link is traversed, for example when opening an object through it. An attacker who can get a user or application to access a crafted file can cause a denial of service, and adjacent heap contents may be incorporated into the external file name or object path the library attempts to open.
Références et Patchs
Dernières Vulnérabilités
CVE-2026-82976
A flaw was found in the image-customization-controller used in OpenShift's bare metal provisioning stack. The controller registers an HTTP file server on port 8084 with no authentication. The custom imageFileSystem implementation exposes a browsable HTML directory listing of all registered images at the root path. The only access control on per-host images is the random UUID filename; the directory listing discloses those UUIDs to any caller on the provisioning network. Each served image embeds an Ignition config containing the base64-encoded cluster pull secret, optional IRONIC_RAMDISK_SSH_KEY, proxy settings, and per-host NetworkManager keyfiles. An adjacent network attacker can enumerate all image UUIDs via directory listing, download any image, extract the embedded Ignition config, and recover the cluster pull secret and SSH key. ### Evidence **pkg/imagehandler/imagefilesystem.go — root directory is browsable and enumerates all UUID image names** ```go func (f *imageFileSystem) Readdir(n int) ([]fs.FileInfo, error) { f.mu.Lock() defer f.mu.Unlock() result := []fs.FileInfo{} for _, im := range f.images { result = append(result, im) } return result, nil } func (f *imageFileSystem) Open(name string) (http.File, error) { f.log.Info("Open", "path", name) if name == "/" { return f, nil } ... ``` **cmd/controller/main.go — file server bound on all interfaces without auth or TLS** ```go http.Handle("/", http.FileServer(imageServer.FileSystem())) go func() { server := &http.Server{ Addr: imagesBindAddr, // default ":8084" ReadHeaderTimeout: 5 * time.Second, } err := server.ListenAndServe() ``` **pkg/ignition/service_config.go:94-100 — pull secret is embedded in every served image** ```go func (b *ignitionBuilder) authFile() ignition_config_types_32.File { source := "data:;base64," + strings.TrimSpace(b.ironicAgentPullSecret) return ignition_config_types_32.File{ Node: ignition_config_types_32.Node{Path: "/etc/authfile.json"}, ... ``` ### Attack Pattern Adjacent provisioning-network client → GET / lists UUID image names → download ISO/initramfs → extract embedded Ignition → recover cluster pull secret and SSH key.
CVE-2026-63417
Logging Redactor is a Python library designed to redact sensitive data in logs based on regex patterns and / or dictionary keys. Prior to 0.0.7, personally identifiable information returned by custom object __str__ or __repr__ methods is not redacted when such non-primitive objects are logged, causing sensitive values to appear in logs that are expected to mask them. This issue is fixed in version 0.0.7.
CVE-2026-63198
traQ is a messenger application built for Digital Creators Club traP. Prior to 3.28.3, OAuth2 silent authorization requests using prompt=none with an empty or omitted scope parameter check prior consent against the empty requested scope set even though all scopes registered for the client become eligible for issuance. An attacker who controls an OAuth client capable of receiving authorization codes can target a user who previously consented to only some client scopes and obtain an access token containing broader registered scopes without additional consent. A user who has never consented to the client still receives consent_required. This issue is fixed in version 3.28.3.