Denial of service in the ip-address npm library before 10.7.4 allows unauthenticated remote attackers to consume CPU and stall applications that pass untrusted input containing many '%' characters into the Address6 IPv6 constructor. The flaw is CWE-1333 inefficient regular expression / quadratic complexity, rated medium severity (CVSS 4.0 base 6.3), with availability-only partial impact; exploitation depends on a reachable application code path that forwards attacker-controlled data to Address6 parsing (CVSS AT:P). No public exploit code was identified at time of analysis, and upgrading to 10.7.4 or later removes the condition.
Gitea before 28.2.0 allows a user with write access to a repository's releases to link an existing issue or comment attachment into a release by supplying its UUID, because the release attachment-linking routine validated the repository and release association but not whether the attachment was already bound to an issue or comment. This exposes the attachment's file name, size and download count and lets the attacker rename or delete it. The flaw is limited to the same repository and requires knowledge of a non-enumerable UUID, and no public exploit or active exploitation has been identified at time of analysis (EPSS ~0.15%).
Out-of-bounds heap reads, and under a non-default configuration out-of-bounds heap writes, in librtmp as shipped with RTMPDump through 2.6 let an unauthenticated attacker who controls or intercepts an RTMP server corrupt or leak memory inside the connecting client. AMF3_Decode ignores failures returned by the AMF3 property decoder, so an oversized AMF3 string length supplied by the server is trusted and used against buffers it does not actually describe, producing an over-read of heap bytes; the write primitive is only reached when the client is invoked with the --token option that enables the DecodeTEA path. Scored 6.1, with no public exploit identified at time of analysis, and real exposure confined to hosts that actively act as RTMP clients against attacker-reachable servers.
Authorization bypass in ProjectSend 2.0.0 through 2.6.x lets authenticated staff holding the import_orphans permission delete or adopt files owned by other users by submitting non-canonical path aliases such as './a/b.txt' to the orphan import or delete endpoints. The root cause is an incorrect-order validation flaw in OrphanFileScanner::isOrphan() (CWE-180): path strings are compared before Flysystem normalization, so a spelling that storage rewrites to a real tracked path is misclassified as an orphan and the delete_others_files boundary is bypassed. The issue is fixed in ProjectSend 2.7.0; exploitation requires a staff account granted import_orphans and the presence of other users' files, and no public exploit was identified at time of analysis.
Stack-based buffer overflow in libstrophe through 0.14.0 allows a malicious or compromised XMPP server to crash or corrupt the stack of a connecting client during SCRAM SASL authentication: a server-first-message whose base64 s= (salt) attribute decodes to more than 124 bytes overruns the fixed 128-byte stack buffer in SCRAM_Hi(). Only clients that perform SCRAM authentication and connect out to an attacker-controlled or compromised server are exposed, so the flaw is client-side and requires the victim to initiate the connection rather than exposing an attack surface at scale. Impact is predominantly denial of service (process abort), with limited stack memory overwrite possible only in NDEBUG builds where the guarding assertion is compiled out; no public exploit code or confirmed active exploitation (CISA KEV) is identified at time of analysis, though a technical write-up of the flaw is publicly available.
Stack-based out-of-bounds write in lwIP's mDNS response handler affects lwIP through 2.2.1 when the optional mDNS application is compiled in and the device is performing an mDNS service search. An unauthenticated attacker on the same local link can send a multicast PTR answer with an oversized RDLENGTH or an SRV answer with RDLENGTH below 6, corrupting a 256-byte stack buffer and causing a crash with possible limited integrity impact. Patch guidance is available via an upstream commit, no public exploit identified at time of analysis, and no CISA KEV entry is present; the assessed severity is moderate (CVSS 6.0).
Account takeover in Gitea 0 through 28.1.0 via the OpenID association callback, which accepts a positive OpenID 2.0 assertion without verifying that the signed-in victim initiated the linking. A remote unauthenticated attacker who controls an OpenID provider can trick a signed-in user into opening a crafted link, link the attacker's identity to the victim's account, and then sign in as the victim through OpenID login, bypassing local TOTP/WebAuthn checks. Exploitation requires OpenID login/association to be enabled and user interaction; no public exploit code has been identified, EPSS is 0.11% (1st percentile), and the vendor has released Gitea 28.2.0 to fix it.
Heap out-of-bounds read in librtmp (RTMPDump) through 2.6 can crash RTMP clients when AMFProp_Decode parses an AMF0 strict-array element count without first confirming that four bytes remain. The issue is unauthenticated from the network perspective but requires the victim to initiate an RTMP session with an attacker-controlled server or for an attacker to be positioned as a man-in-the-middle; impact is limited to availability (client crash), with no code execution or confidentiality impact established. No public exploit code was identified at time of analysis.
Cross-site request forgery in Dyad's OAuth Callback Handler affects Dyad through version 1.16.0 and can let a crafted dyad:// callback be accepted without proof that it belongs to a legitimate in-app connection flow, causing an integrity-only credential/integration state change. Exploitation is remote but requires the Dyad desktop app to be installed with its custom-protocol handler registered and requires the victim to launch or approve the crafted protocol link; there is no confidentiality or availability impact per the independent assessment, and the CVSS 4.0 base score is 5.3. Vendor-released patch: 1.17.0-beta.1; no public exploit identified at time of analysis.
Unauthenticated requests to Gitea's organization home route can trigger a nil-pointer panic and HTTP 500 when the path segment resolves to an individual user instead of an organization. The issue affects Gitea through 28.1.0 and is fixed in 28.2.0; it requires no authentication or special configuration, but each request causes only per-request availability degradation (error response and logged panic), not a full service crash. EPSS is 0.19% (8th percentile) and no public exploit has been identified at time of analysis.
Incorrect authorization in OneUptime 14.0.2 through 14.0.68 allows an authenticated project member holding the AI-access admin-set permission to act on Kubernetes clusters that their label- and owner-based scope does not cover, including probing agent RBAC permissions via the AI access test, resetting/disconnecting agents, approving pending kubectl changes, and steering incident AI conversations into running commands on unauthorized resources. The flaw is reachable only where cluster management and AI agent features are actually in use, and the incident-conversation path is limited to read-only kubectl/infrastructure commands; no public exploit code was identified at time of analysis and the issue is not in CISA KEV. Vendor-released patch: 14.0.69, with a documented regression test suite exercising per-cluster edit scope through the real user middleware.
Stored cross-site scripting in the Bootstrap Shortcodes plugin for WordPress before 3.4.6 allows authenticated users with Contributor-level access or higher to inject arbitrary web script through a crafted id attribute in the [bs_tcontent] shortcode, which is rendered into an unquoted HTML attribute; a victim must view the page and interact with the injected element for the script to execute. The issue is rated CVSS 5.1 and there is no public exploit identified at time of analysis; the vendor has released version 3.4.6 with a fix that quotes the attribute.
Signed-in Gitea users can bypass the interaction restriction of a locked issue by adding or removing emoji reactions on the issue and its comments through the web UI, even though they lack write access and the REST API correctly refuses the same operation. The gap affects Gitea through 28.1.0 and is limited to reaction toggling - no issue text, comment content, labels, or repository state can be altered - so the practical consequence is cosmetic noise and notification churn on issues maintainers deliberately froze. There is no public exploit identified at time of analysis, CISA SSVC records no known exploitation, and EPSS is 0.18% (7th percentile), consistent with a low-severity authorization inconsistency rather than a high-impact access-control failure; the vendor-released fix is Gitea 28.2.0.
Information disclosure in Gitea 1.22.0 through 28.1.0 allows an authenticated low-privileged user who can create a draft release to obtain attachment metadata from repositories they cannot read, as well as metadata for abandoned uploads. The draft-release API path builds an assets list from an unsaved release whose ID is 0, causing the attachment lookup to return file name, UUID, upload time, size, and download count, but not attachment contents. CISA SSVC reports exploitation: none, no public exploit was identified at time of analysis, and EPSS is 0.16% (4th percentile).
Gitea leaks private repository metadata to users whose access has since been revoked: the GET /api/v1/user/subscriptions and /api/v1/user/starred endpoints returned name, description, clone URLs, owner and counters for every watched or starred private repo without re-checking current permissions, so a user who starred a private repo while authorized and later lost access (for example via a team permission downgrade or collaborator removal) could keep reading that metadata with an otherwise valid account. The flaw is fixed in 1.27.0 and 28.0.0 and was disclosed with 28.2.0; EPSS is only 0.15% (4th percentile), there is no KEV listing and no public exploit code identified at time of analysis, and the CVSS 4.3 vector (AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N) reflects that an attacker must already be authenticated and must previously have held legitimate access, with only metadata - not repository contents - exposed.