Just published.
The newest CVEs, straight from CVE.org and NVD.
-
CVE-2026-88215
Use after free in Windows Filtering Platform (WFP) allows an authorized attacker to elevate privileges locally.
-
CVE-2026-88216
Time-of-check time-of-use (toctou) race condition in Windows DHCP Server allows an authorized attacker to execute code over a network.
-
CVE-2026-88217
Use of uninitialized resource in Windows Graphics Kernel allows an authorized attacker to disclose information locally.
-
CVE-2026-88218
Double free in Windows DNS allows an unauthorized attacker to execute code over a network.
-
CVE-2026-88213
Numeric truncation error in Windows Bluetooth Service allows an authorized attacker to elevate privileges locally.
-
CVE-2026-54436
Heap-based Buffer Overflow in the io-char library's DCMD_CHR_WAITINFO devctl() handler could allow an attacker with local access to cause data tampering, arbitrary code execution, or a crash in the Character I/O device driver.
-
CVE-2026-54437
Improper Input Validation in the io-char library's DCMD_CHR_TCINJECTC devctl() handler could allow an attacker with local access to cause data tampering in the memory input buffer of the Character I/O device driver.
-
CVE-2026-40275
Heap-based Buffer Overflow in the DCMD_PROC_EVENT handler of the procnto could allow an attacker with local access to cause an information disclosure, data tampering or a crash of the QNX Neutrino kernel.
-
CVE-2026-40274
Invalid Pointer Reference in the TimerSettime() system call could allow an attacker with local access to arm a crafted timer event which can cause an information disclosure in the QNX Neutrino kernel.
-
CVE-2026-18448
EDK2 contains a vulnerability in BIOS where a privileged attacker may cause “Exposure of Sensitive Information during Transient Execution” by local access. Successful exploitation of this vulnerability may result in disclosure of SMRAM contents. It could also impact Confidentiality.
-
CVE-2026-40273
Improper Input Validation in a procnto message handler could allow an attacker with local access to send a crafted message which can cause an information disclosure in the QNX Neutrino kernel
-
CVE-2026-14968
EDK2 contains vulnerabilities in BIOS where skilled and privileged users or attackers may cause “Out-of-bounds Write” by local access. Successful exploitation of this vulnerability could compromise SMM handler integrity, cause system instability, and lead to denial of service. It could also impact Integrity and Availability.
-
CVE-2026-126134
Thunderbird allowed feed subscriptions to file: URLs that point to a network share. If a user was tricked into subscribing to such a URL, Thunderbird connected to the attacker's server when fetching the feed. Windows then sent the user's NTLM credentials to that server. This bug only affects Thunderbird on Windows. Other operating systems are unaffected. This vulnerability was fixed in Thunderbird 160 and Thunderbird 153.7.
-
CVE-2026-126135
If an attacker was able to perform a man-in-the-middle attack on an SMTP or POP3 server connection using STARTTLS, the attacker could inject server responses before the TLS handshake completed. Thunderbird would process these responses after the handshake, as if they had been received from the server over the encrypted connection. This could be used to make Thunderbird report a different outcome for sending or fetching mail, and to show attacker-controlled text in error messages. The encrypted connection itself was not compromised, and credentials were not exposed. Connections using SSL/TLS instead of STARTTLS were not affected. This vulnerability was fixed in Thunderbird 160 and Thunderbird 153.7.
-
CVE-2026-126136
For Exchange accounts using EWS or Microsoft Graph with OAuth2 authentication, Thunderbird sent the OAuth2 access token to the server even when the configured server URL used unencrypted HTTP. An attacker able to observe network traffic could obtain the token and use it to access the user's mailbox until the token expired. An attacker positioned on the network during account setup could cause such a URL to be configured. This vulnerability was fixed in Thunderbird 160 and Thunderbird 153.7.
-
CVE-2026-126087
Untrusted search path in the cross-user process launcher in Open Job Description (OpenJD) Sessions for Python before 0.11.0, as used in the AWS Deadline Cloud worker agent before 0.33.0, on Linux might allow an authenticated remote user with permission to submit jobs to a queue to execute arbitrary code as the worker agent OS user and obtain the AWS credentials held by the worker agent via a crafted job that modifies the PATH environment variable. To remediate this issue, users of customer-managed fleets should upgrade the Deadline Cloud worker agent to version 0.33.0 or later, which includes OpenJD Sessions 0.11.0. Service-managed fleets were remediated service-side.
-
CVE-2026-126132
When a message was forwarded inline, edited as new, or had an attachment dragged into a compose window, a specially crafted link attachment (a .desktop file on Linux, or a .url or .lnk file on Windows) was replaced with the content of the file or URL it pointed to when the message was saved or sent. This could disclose local files, or content from the user's network, to the message recipients or to the mail provider. This bug only affects Thunderbird on Linux and Windows. Other operating systems are unaffected. This vulnerability was fixed in Thunderbird 160 and Thunderbird 153.7.
-
CVE-2026-126133
When a CalDAV or ICS calendar was accessed through an HTTP proxy that requested authentication, Thunderbird could answer the proxy with the calendar server's username and password. This affected both setting up a new calendar and syncing existing calendars, where a saved password could be sent without any user interaction. A malicious or compromised proxy could use this to obtain the calendar credentials, even for calendars served over HTTPS. This vulnerability was fixed in Thunderbird 160 and Thunderbird 153.7.
-
CVE-2026-125334
Forge (also called node-forge) is a native implementation of Transport Layer Security in JavaScript. Prior to 1.4.3, a forge.tls client can pass a handshake record to tls.handleHandshake() while the connection has no session, including after a close_notify alert, after the application calls close(), or before handshake() is called. The handler dereferences c.session.md5 before dispatching the message, causing a TypeError to propagate from connection.process(). A malicious server or on-path attacker can deliver the record, including after a reset when the connection uses no encryption, and terminate a Node.js process if the application does not catch the exception. forge.tls servers are not affected, and there is no confidentiality or integrity impact. This issue is fixed in version 1.4.3.
-
CVE-2026-125335
Forge (also called node-forge) is a native implementation of Transport Layer Security in JavaScript. Prior to 1.4.3, ed25519.verify() decodes a public key in crypto_sign_open() without rejecting small-order points or their non-canonical encodings. With S = 0 and a small-order R value, an attacker who can choose the public key can produce signatures that verify for many messages, or all messages with the identity key, without a private key. This can defeat proof-of-possession, attribution, non-repudiation, or deduplication checks that rely on a signature being bound to the message. Keys produced by normal Ed25519 key generation are not small-order, so applications that verify only trusted or internally generated public keys are not affected. This issue is fixed in version 1.4.3.
-
CVE-2026-125336
Forge (also called node-forge) is a native implementation of Transport Layer Security in JavaScript. Prior to 1.4.3, pki.verifyCertificateChain() in lib/x509.js does not decode or enforce the X.509 nameConstraints extension while validating certificate paths. A name-constrained certificate authority with a non-critical extension can therefore issue a leaf certificate with a subjectAltName dNSName or subject commonName outside a permitted subtree or inside an excluded subtree, and node-forge accepts the chain. Exploitation requires a name-constrained CA that chains to a trusted root and either its private key or the ability to obtain an out-of-scope certificate, allowing impersonation of names outside the delegated namespace, including against forge.tls clients. Chains without nameConstraints are not affected, and affected versions safely reject critical nameConstraints as unsupported rather than accepting them. This issue is fixed in version 1.4.3.
-
CVE-2026-125328
pgEdge Go LLM Library provides a unified Go interface for multiple LLM providers with streaming, tool calling, retries, and an HTTP and server-sent events proxy. From 0.1.0 until 0.2.0, GET /v1/health in llm/proxy/handlers.go uses handleHealth to ping providers with the operator's real credentials and returns their error text, while the Authorize hook in llm/proxy/proxy.go permits access by default when the consuming application does not configure it. When a provider rejects the configured credential, a remote caller who can reach the proxy can obtain part of the operator's third-party provider API key from the resulting error. This issue is fixed in version 0.2.0.
-
CVE-2026-125329
Kaneo is a project management application with boards, Gantt charts, and GitHub integration that can be self-hosted or used as a cloud service. Prior to 2.13.0, GET /api/external-link/task/:taskId returns integration.config to any workspace member, including the viewer role, when a task has a linked Gitea issue or pull request. The plaintext integration.config contains the Gitea accessToken and webhookSecret, allowing the token's granted Gitea actions and forged webhook deliveries that pass HMAC verification. GitHub integrations and installations without a configured Gitea integration are not affected. This issue is fixed in version 2.13.0.
-
CVE-2026-125332
Forge (also called node-forge) is a native implementation of Transport Layer Security in JavaScript. Prior to 1.4.3, a forge.tls client accepts a ServerHello compression_method that it did not offer. When the client later handles ServerHelloDone, tls.createConnectionState() rejects the unsupported compression algorithm by throwing an Error that propagates from connection.process(). A malicious server or an on-path attacker can alter the unauthenticated ServerHello compression byte before the server Finished message and terminate a Node.js process when the application does not catch the exception. The affected client paths include direct forge.tls use, forge.tls.wrapSocket, and forge.http; forge.tls servers are not affected, and the issue has no confidentiality or integrity impact. This issue is fixed in version 1.4.3.
-
CVE-2026-125330
HarnessRouter is a self-hosted unified API for agent harnesses. From version 0.4.2 until 0.31.7, routes that accept a session ID or response ID validate only the caller's organization and do not enforce the workspace scope of a non-default workspace API key. A key holder who learns a non-guessable identifier from another workspace in the same organization can read sessions, conversation turns, files, traces, and responses, overwrite files, cancel or delete sessions, and publish share links. The issue does not cross organization boundaries, and console users, organization keys, and Default Workspace keys intentionally retain organization-wide access. This issue is fixed in version 0.31.7.
-
CVE-2026-125331
HarnessRouter is a self-hosted unified API for running agent harnesses with sessions, streaming, files, cancellation, and failure handling. Prior to 0.31.8, a console that is not in self-hosted mode attaches the server's internal key to requests carrying an Authorization header before validating the bearer, while the service-only PUT and GET /v1/orgs/{org}/connections/{name}, PUT /v1/orgs/{org}/policy/{backend}, and /internal/* routes trust that key and take the organization from the URL. In a multi-organization deployment, a remote caller can use a made-up bearer to rewrite another organization's provider connection to an attacker-controlled endpoint, causing later prompts to be disclosed and attacker-controlled responses to influence tasks. This issue is fixed in version 0.31.8.
-
CVE-2026-125333
Forge (also called node-forge) is a native implementation of Transport Layer Security in JavaScript. From 0.1.2 until 1.4.3, forge.http uses _trimString() in lib/http.js to remove trailing whitespace from response header values and cookie names and values with a regular expression that can take quadratic time. A malicious server or on-path attacker can supply a long whitespace run followed by a non-whitespace character in a response header, causing response.readHeader() or response.getCookies() to repeatedly scan the input and block the Node.js event loop or browser tab for minutes. Applications are affected only when they use forge.http with a socket pool to communicate with an untrusted server; forge.http is not loaded by require('node-forge'), and the issue has no confidentiality or integrity impact. This issue is fixed in version 1.4.3.
-
CVE-2026-125327
pgEdge RAG Server is an API server for retrieval-augmented generation over PostgreSQL content using pgvector. Prior to 2.0.0, the unauthenticated query endpoint uses FetchDocuments in internal/database/search.go to read the entire filtered corpus without a LIMIT and rebuild an in-memory BM25 index on every request. Because HybridEnabled is set by default in internal/config/loader.go and the endpoint has no rate limit, inexpensive remote queries can impose database and CPU work proportional to corpus size and exhaust service resources. This issue is fixed in version 2.0.0.
-
CVE-2026-123387
Improper neutralization of special elements used in a command ('command injection') in Windows Active Directory allows an authorized attacker to elevate privileges over a network.
-
CVE-2026-123388
Heap-based buffer overflow in Windows Program Compatibility Assistant Service allows an authorized attacker to elevate privileges locally.
-
CVE-2026-123389
Use after free in Quality Windows Audio/Video Experience (QWAVE) service allows an authorized attacker to elevate privileges locally.
-
CVE-2026-122984
Heap-based buffer overflow in Windows Universal Disk Format File System Driver (UDFS) allows an authorized attacker to elevate privileges locally.
-
CVE-2026-122297
Use of uninitialized resource in Kernel Transaction Manager allows an authorized attacker to disclose information locally.
-
CVE-2026-122474
Time-of-check time-of-use (toctou) race condition in Windows Security Account Manager allows an unauthorized attacker to bypass a security feature over a network.
-
CVE-2026-122475
Untrusted pointer dereference in Windows Storage Spaces Controller allows an unauthorized attacker to execute code over a network.
-
CVE-2026-122174
Improper link resolution before file access ('link following') in Windows Task Scheduler allows an authorized attacker to elevate privileges locally.
-
CVE-2026-122176
Heap-based buffer overflow in Windows SMB allows an authorized attacker to execute code over a network.
-
CVE-2026-122177
Improper link resolution before file access ('link following') in Windows iSCSI allows an unauthorized attacker to elevate privileges over a network.
-
CVE-2026-122169
Insufficient granularity of access control in Windows IKE Extension allows an authorized attacker to perform spoofing locally.
-
CVE-2026-122170
Use after free in Windows Win32K allows an authorized attacker to elevate privileges locally.
-
CVE-2026-122173
Heap-based buffer overflow in HID Parser Library allows an unauthorized attacker to elevate privileges with a physical attack.
-
CVE-2026-122164
Weak authentication in Windows Kerberos allows an authorized attacker to elevate privileges over a network.
-
CVE-2026-122165
Heap-based buffer overflow in Windows Client for NFS allows an authorized attacker to elevate privileges locally.
-
CVE-2026-122167
Heap-based buffer overflow in Windows USB Driver allows an unauthorized attacker to elevate privileges with a physical attack.
-
CVE-2026-122168
Heap-based buffer overflow in Microsoft Windows Media Foundation allows an authorized attacker to elevate privileges locally.
-
CVE-2026-122163
Double free in Windows Image Acquisition allows an unauthorized attacker to execute code over a network.
-
CVE-2026-115559
IMAP4.starttls() wraps the socket in TLS but leaves self._readbuf, imaplib's own read-ahead buffer, untouched. Any plaintext bytes that arrived in the same recv() as the tagged OK to STARTTLS are returned by readline() after the handshake, so the client treats them as the first responses on the protected channel. A network attacker who cannot break TLS can append forged lines after the server's OK, let the handshake complete with the real server and a valid certificate, and have the client accept the forged CAPABILITY and the forged tagged completion of the next command. This is the STARTTLS command/response injection class of CVE-2011-0411.
-
CVE-2026-107496
Issue summary: Encrypting a CMS message to an EC certificate whose group order is 1 may cause CMS_encrypt() or CMS_encrypt_ex() to loop indefinitely. Impact summary: An attacker able to supply the recipient certificate to an application performing CMS encryption can cause that application to loop indefinitely, consuming CPU and resulting in Denial of Service. CWE: CWE-835: Loop with Unreachable Exit Condition ('Infinite Loop') Description: CMS_encrypt() and CMS_encrypt_ex() encrypt a message to one or more recipient certificates and, for an EC recipient, generate an ephemeral key in the recipient's group. If the certificate declares a group order of 1, random scalar generation can only produce zero. The key generation routine then retries without limit while waiting for a nonzero private scalar, so the call never returns. The OpenSSL TLS and QUIC implementations do not use this path. An application is exposed only if it calls CMS_encrypt() or CMS_encrypt_ex() with a recipient certificate whose EC parameters are attacker-controlled. Certificates with normal EC group orders do not trigger the loop. OpenSSL 4.0 and later reject unknown explicit EC groups in default builds. These versions are affected only when configured with the enable-ec_explicit_curves option. The result is a hang that consumes a CPU core until the process is interrupted, a Denial of Service. The severity is Low because the attacker must supply the recipient certificate and the application must choose to encrypt to it; ordinary CMS encryption to trusted recipients is unaffected. The fix makes EC key generation fail when the group has no valid nonzero private scalar, instead of retrying indefinitely. FIPS impact: no The FIPS modules in 4.1, 4.0, 3.5 and 3.0 are not affected. The FIPS provider does not support arbitrary explicit EC domain parameters, so an order-1 group is rejected during key import before the vulnerable key generation path is reached.
-
CVE-2026-105437
Affected versions of Git have a heap buffer overflow in the three-way merge code (xdiff), which an attacker can trigger through a repository they control. When Git merges a file, it first adds up the size of the merged output, allocates a buffer of that size, and then writes the output into it. That total was kept in a 32-bit signed integer. The conflict-marker-size attribute, which can be set from a committed .gitattributes file, accepts any positive integer, and the markers are written several times for every conflict. A small file with a huge marker size and many separate conflicts can therefore push the total past 4 GiB, so the integer wraps to a small value and Git writes attacker-controlled merge output past the end of a buffer that is too small. At minimum this crashes Git, and because the overflow is large and its contents are largely attacker-controlled, worse heap corruption is possible. Any command that does a textual three-way merge can trigger it, such as git-merge(1), git-rebase(1), git-cherry-pick(1), and git-merge-tree(1). The latter is most likely a target for Git forges, which might use git-merge-tree(1) on the server-side to check mergeability. When the server reads attributes from trees in the repository (for example through attr.tree or --attr-source), any user who can push branches and request that they be merged (for example by opening a pull or merge request) can trigger the overflow without anyone else doing anything, and the data they push is small because the oversized output only appears during the merge. The problem has been patched by computing the size as size_t and checking every addition for overflow, so Git stops with an error instead of overflowing. Servers that cannot upgrade right away can add * conflict-marker-size=7 to each repository's info/attributes as a workaround, which overrides any in-tree .gitattributes, or turn off server-side mergeability checks for untrusted repositories.
-
CVE-2026-120003
Heap-based buffer overflow in Microsoft Office allows an unauthorized attacker to execute code over a network.
The newest published CVE records from CVE.org and NVD, refreshed every few minutes, with severity, exploit odds and fix status.