These CVE IDs are still marked RESERVED at MITRE — no official description, no CVSS, no NVD record — yet the world is already talking about them. Mallory tracks the chatter so you see the risk before the paperwork catches up.
1,178 reserved CVEs with public mentions, ranked by all-time mention count.
Page 44 of 48
CVE-2026-49455 is a cross-origin request forgery vulnerability in Waku's React Server Components (RSC) server action dispatch mechanism. The issue allows authenticated cross-origin POST requests to reach registered "use server" action endpoints without proper Origin validation. As a result, a malicious site can induce a victim's browser to submit POST requests to Waku server action endpoints while automatically including the victim's session cookies, causing server-side actions to execute in the victim's authenticated context. The vulnerable condition is specifically the absence of effective Origin checking on POST requests to the server action dispatch path, including deployments using the configured rscBase prefix.
CVE-2026-49455First seen Jul 9, 2026
CVE-2026-49831 is a path traversal vulnerability in DSpace affecting the reporter output path used by Curation Tasks. According to the provided advisory, an authenticated DSpace Collection, Community, or Site Administrator can influence the reporter output location and cause reporter output to be written to unintended writable filesystem paths. This can allow overwriting files outside the expected directory scope, including locations such as /dspace/config or /dspace/bin. The issue is therefore consistent with improper limitation of a pathname to a restricted directory.
CVE-2026-49831First seen Jul 9, 2026
CVE-2026-54333 is a stack-based memory corruption vulnerability in the uefi-firmware parser’s native Tiano/EFI decompressor. According to the provided advisory, the flaw is located in the MakeTable function, where processing a crafted firmware blob can cause a stack out-of-bounds write. The issue affects uefi-firmware versions up to and including 1.12. Because the vulnerable code operates while parsing firmware content, an attacker able to supply a maliciously formed firmware image or blob to the parser can trigger corruption of stack memory during decompression-related table construction.
CVE-2026-54333First seen Jul 9, 2026
DSpace contains a possible remote code execution vulnerability in its handling of Velocity templates used for COAR Notify/Linked Data Notifications (LDN) messages. According to the provided advisory, an attacker with DSpace administrator credentials may be able to execute Java code directly from Velocity templates via reflection. The issue is especially dangerous when chained with the related LDN path traversal issue, which can expand reachable attack paths. The vulnerable condition is tied to template processing in the COAR Notify/LDN functionality rather than general unauthenticated request handling.
CVE-2026-49832First seen Jul 9, 2026
CVE-2026-49833 is a path traversal vulnerability in DSpace's COAR Notify / Linked Data Notifications (LDN) message generation. The flaw allows an authenticated DSpace administrator to reference files outside the intended trusted template directory during LDN template handling. Because the referenced file is processed as an Apache Velocity template, the issue can lead not only to arbitrary file read and information disclosure, but also, in demonstrated attack chains, arbitrary Java code execution through unsafe interpretation of attacker-controlled or attacker-selected template content by the Velocity engine.
CVE-2026-49833First seen Jul 9, 2026
EGroupware contains a local file inclusion issue in its mail compose functionality due to unsafe handling of URI schemes embedded in HTML email bodies. According to the provided advisory, attacker-controlled HTML content can include inline image or similar resource references using file:// URIs, and the application subsequently retrieves the referenced resource via file_get_contents() without enforcing a strict HTTP/HTTPS-only allowlist. This allows an authenticated user with access to the mail feature to cause the server to read arbitrary local files accessible to the web server process and include their contents as inline MIME attachments in outgoing email.
CVE-2026-45016First seen Jul 8, 2026
CVE-2026-53508 affects github.com/oasdiff/oasdiff. The vulnerability is a policy-enforcement bypass in the git-revision input/load path (rev:path): oasdiff does not correctly enforce the --allow-external-refs=false setting when loading OpenAPI specifications through that path. As a result, external $ref targets can still be resolved even though external references were explicitly disabled. In practice, when oasdiff processes an untrusted OpenAPI document via the git-revision load path, an attacker can embed external $ref values pointing to HTTP(S) resources or local filesystem paths, including file:// URIs, and cause the tool to dereference them. This creates an SSRF and local file read condition despite the intended safeguard.
CVE-2026-53508First seen Jul 8, 2026
CVE-2026-53553 is a path traversal vulnerability in Goploy affecting the /deploy/fileDiff endpoint. According to the provided advisory, an authenticated low-privileged user can supply a crafted filePath value containing traversal sequences to escape intended directory boundaries. The flaw affects both local file access on the Goploy host and remote file access against managed servers reached via SFTP. Because path input is not properly validated, sanitized, canonicalized, and restricted to an approved base directory before file operations occur, an attacker can cause the application to read arbitrary files from the local host and from configured remote systems.
CVE-2026-53553First seen Jul 8, 2026
CVE-2026-34151 is a path traversal vulnerability in XWiki Platform Old Core affecting the /skin/ action endpoint when XWiki is deployed on Jetty 12 or later. The flaw allows crafted request paths to traverse directories and access arbitrary resources that are readable by the Jetty process. According to the advisory, this can expose sensitive files such as /etc/passwd as well as XWiki configuration files located under WEB-INF. The issue is deployment-specific and is described as affecting Jetty 12+; Tomcat and earlier Jetty versions are stated as not impacted.
CVE-2026-34151First seen Jul 8, 2026
CVE-2026-55630 is a stored cross-site scripting vulnerability in Kiwi TCMS affecting the TestCase.extra_link and TestPlan.extra_link fields. According to the provided advisory, these fields accepted JavaScript: URI values without proper sanitization, allowing attacker-supplied script payloads to be stored and later executed in a victim's browser when the malicious link is rendered or followed in the application UI. The issue was fixed in Kiwi TCMS 16.1 by sanitizing input for these fields and resetting existing invalid records in the database to null.
CVE-2026-55630First seen Jul 7, 2026
CVE-2026-54641 is an authorization flaw in io.openremote:openremote-manager affecting multi-tenant realm isolation. According to the provided advisory, three read methods in UserResourceImpl—get, getUserClientRoles, and getUserRealmRoles—do not properly enforce realm-access authorization before returning user information. As a result, a realm administrator in one tenant with read:admin privileges can query user records belonging to other realms, including the master realm. The exposed data includes user profile information, account enabled/disabled status, and assigned Keycloak client and realm roles.
CVE-2026-54641First seen Jul 7, 2026
CVE-2026-54640 is an XML External Entity (XXE) vulnerability in OpenRemote's openremote-agent, specifically in the KNXProtocol asset import handling path. According to the provided advisory, this is an incomplete fix related to CVE-2026-40882: the KNXProtocol asset import handler, including KNXProtocol.startAssetImport(), remains vulnerable because the XML parsing path does not fully disable DTD processing and external entity resolution. The advisory specifically notes exposure in both Saxon/XSLT processing and XMLInputFactory usage. A successful attacker can supply crafted XML during asset import to cause the parser to resolve external entities, enabling arbitrary file disclosure and potentially server-side request forgery (SSRF).
CVE-2026-54640First seen Jul 7, 2026
flyto-core contains an unauthenticated command execution vulnerability in the HTTP MCP POST /mcp endpoint. The issue is exposed through tools/call execute_module, which allows invocation of sandbox.execute_shell and results in arbitrary OS command execution by the flyto-core server process. The advisory indicates the vulnerable request handling does not properly enforce authentication on /mcp routes and does not adequately apply module filtering/denylisting to prevent access to dangerous sandbox functionality.
CVE-2026-55786First seen Jul 7, 2026
OpenRemote Manager contains an authorization flaw in the predicted datapoint write endpoint. According to the provided advisory, users granted only read:assets privileges can still write predicted datapoints for assets, indicating that the endpoint fails to properly enforce write-level authorization checks. This results in a privilege bypass where a nominally read-only asset user can perform unauthorized modification operations against predicted datapoints.
CVE-2026-49439First seen Jul 7, 2026
CVE-2026-54637 is an unauthenticated server-side request forgery (SSRF) vulnerability in Dragonfly scheduler v1 and v2. According to the provided advisory, the issue is in the scheduler v1 gRPC DownloadTinyFile path, where attacker-controlled PeerHost values can cause the scheduler to initiate server-side HTTP GET requests to arbitrary internal addresses. The response data from those requests can then be stored in Task.DirectPiece. The vulnerable logic is associated with scheduler/resource/standard/peer.go, and the recommended fix is to use nethttp.NewSafeDialer().DialContext in the HTTP transport for DownloadTinyFile. The advisory also notes that additional validation in storeHost to reject non-global-unicast addresses is advisable.
CVE-2026-54637First seen Jul 7, 2026
CVE-2026-54724 is an open redirect vulnerability in Kiwi TCMS affecting the account confirmation endpoint. The issue is caused by insufficient validation of a user-controllable `next` parameter, which allows an attacker to supply an arbitrary external URL that is used as the post-confirmation redirect target. As a result, a victim following a legitimate-looking Kiwi TCMS account confirmation link can be redirected off-site to attacker-controlled infrastructure. Based on the provided content, affected versions include all versions through 12.4.
CVE-2026-54724First seen Jul 7, 2026
OpenRemote Manager contains an insecure direct object reference (IDOR) / missing object-level authorization flaw in the bulk alarm deletion endpoint. According to the advisory, an authenticated user with alarm write permissions in their own realm can supply arbitrary alarm IDs to the vulnerable bulk removeAlarms functionality and delete alarm records belonging to other realms. The issue is cross-tenant in nature because authorization is not properly enforced against the tenant/realm ownership of the referenced alarm objects.
CVE-2026-57168First seen Jul 3, 2026
CVE-2026-49250 is a denial-of-service vulnerability in @conform-to/dom affecting the parseSubmission functionality when processing FormData or URLSearchParams submissions. According to the provided advisory, the vulnerable behavior is triggered by inputs containing many unique field names. The issue causes excessive synchronous CPU consumption during parsing. The fix description indicates the root cause was inefficient processing that repeatedly looked up values by field name rather than iterating submitted entries directly.
CVE-2026-49250First seen Jul 3, 2026
Dragonfly Manager exposes unauthenticated REST API endpoints at GET /api/v1/oauth and GET /api/v1/oauth/:id that return configured OAuth provider records without requiring authentication. The returned JSON includes sensitive OAuth configuration data such as client_secret, client_id, and redirect URL for configured identity providers including GitHub and Google. This results in an information disclosure vulnerability in the Manager API caused by insufficient access control on the OAuth route group and unsafe serialization of secret material in API responses.
CVE-2026-49254First seen Jul 3, 2026
Kiwi TCMS exposed its /init-db/ database setup/initialization page after the application had already been initialized. As described in the advisory, unauthenticated requests could still reach the database initialization proxy endpoint even after first use. Repeated migration attempts were effectively a no-op once migrations had already been applied, so the issue appears to stem from missing or incomplete access restriction on an installation/setup function that should no longer be reachable post-deployment.
CVE-2026-49292First seen Jun 12, 2026
GeoNetwork contains a reflected cross-site scripting vulnerability in its AngularJS-based error page. Based on the provided advisory, the issue is triggered via client-side template injection in reflected input, allowing attacker-controlled content from a crafted URL to be interpreted in the browser and executed as arbitrary JavaScript. The vulnerable condition affects GeoNetwork deployments prior to the fixed supported releases 4.2.15 and 4.4.10; the 3.x and 4.0.x branches are unmaintained and will not receive a fix.
CVE-2026-39379First seen Jul 2, 2026
GeoNetwork contains an authorization bypass in its Elasticsearch-backed search API. According to the provided advisory, when a search request body omits the query field, the application fails to apply the expected ACL/access-control filtering to indexed metadata records. This logic flaw allows requests to return records that should remain restricted, including group-limited, draft, ownership-restricted, and portal-filtered metadata. The issue is therefore an access-control enforcement failure in the search path rather than a memory-safety or injection flaw.
CVE-2026-46487First seen Jul 2, 2026
OpenAM Community Edition through 16.0.6 contains an improper authorization flaw in its OAuth2 authorization-code grant handling. When the realm-wide codeVerifierEnforced setting is disabled, the token endpoint can redeem a PKCE-protected authorization code without validating the required code_verifier. This breaks the intended PKCE binding between the authorization request and token exchange, allowing an intercepted authorization code to be exchanged despite the absence of the verifier.
CVE-2026-48717First seen Jun 30, 2026
OpenAM Community Edition contains an improper authentication flaw in its OAuth2 private_key_jwt client authentication handling. According to the provided advisory, a registered OAuth2 client can impersonate another client when the target client's key material is published through jwks_uri. Successful exploitation allows the attacker to authenticate as the victim client and mint tokens as that client. The issue can potentially affect clients across any realm hosted by the same OpenAM process.
CVE-2026-47426First seen Jun 30, 2026
CVE-2026-47424 is an authenticated remote code execution vulnerability in OpenAM Community Edition caused by a sandbox escape in Groovy-based server-side scripting. Based on the provided advisory, a user who can create or modify server-side scripts that are executed by OpenAM can break out of the intended scripting restrictions and invoke operating-system commands from within the OpenAM JVM. The issue affects the server-side scripting mechanism rather than a client-side component, and the resulting code execution occurs in the security context of the application server administrator account running OpenAM.
CVE-2026-47424First seen Jun 30, 2026