CVE Feed

Improper protection against voltage and clock glitches vulnerability in Microchip SAMA5D4 allows Hardware Fault Injection. This issue affects SAMA5D4.

Potential security vulnerabilities have been identified in HP Easy Start for macOS, versions prior to 2.16.7.260722. These potential vulnerabilities may lead to escalation of privilege. HP is releasing updates to mitigate these potential vulnerabilities.

Potential security vulnerabilities have been identified in HP Easy Start for macOS, versions prior to 2.16.7.260722. These potential vulnerabilities may lead to escalation of privilege. HP is releasing updates to mitigate these potential vulnerabilities.

Potential security vulnerabilities have been identified in HP Easy Start for macOS, versions prior to 2.16.7.260722. These potential vulnerabilities may lead to escalation of privilege. HP is releasing updates to mitigate these potential vulnerabilities.

The Zephyr virtio driver does not validate the descriptor-chain head id that the virtio device writes into the used ring. In virtio_isr() (drivers/virtio/virtio_common.c), the device-written vq->used->ring[idx].id is used directly as an index into vq->recv_cbs[] and vq->desc[], which are both allocated with exactly vq->num entries. recv_cbs[] holds {cb, opaque} callback entries, and the indexed callback pointer is then invoked as cbe.cb(cbe.opaque, used_len). Because the id is consumed as a 16-bit value with no bound check, a malicious or compromised virtio backend (an untrusted hypervisor, or an untrusted hardware/peer-processor virtio device on a PCI or MMIO transport) can supply an id far beyond vq->num. This causes an out-of-bounds read of a {function pointer, argument} pair from heap memory beyond recv_cbs[], after which the driver calls that attacker-shaped pointer in the guest's interrupt context. No guest privileges or user interaction are required; the backend triggers it by writing the shared used ring and raising the queue interrupt. The result is an arbitrary / attacker-influenced function-pointer call in the Zephyr guest, i.e. a control-flow-hijack primitive that can lead to code execution or, at minimum, a reliable crash. The fix rejects any used-ring id >= vq->num before indexing recv_cbs[]/desc[] or invoking the callback. This affects builds using CONFIG_VIRTIO with the PCI or MMIO transport.

The UMP Stream responder library in lib/midi2/ump_stream_responder.c builds reply packets in a 16-byte struct midi_ump (uint32_t data[4]). The builders make_endpoint_info() and make_function_block_info() populate only the first two words (res.data[0] and res.data[1]) and, before this fix, declared their result as an uninitialised local (struct midi_ump res;). The remaining two words (res.data[2], res.data[3]) retain stale stack contents. Endpoint Info and Function Block Info notifications are UMP Stream messages (UMP_MT_UMP_STREAM), which are 4 words long, so the full 16-byte packet — including the two uninitialised words — is transmitted verbatim by cfg->send(). The responder is driven by attacker-supplied UMP Stream Endpoint-Discovery / Function-Block-Discovery requests via ump_stream_respond(). In the in-tree Network MIDI 2.0 server (subsys/net/lib/midi2/netmidi2.c) these requests arrive as UDP datagrams and, with the default no-authentication endpoint, a remote peer can establish a session and trigger the responses; the same library also serves USB MIDI 2.0 hosts. Each discovery request causes the device to disclose 8 bytes of its own uninitialised stack memory to the peer, and the request is freely repeatable. This is a confidentiality-only information leak (root cause is use of an uninitialised variable, CWE-457/CWE-908); the leaked words could include residual data or pointer values. There is no memory-corruption, integrity, or availability impact. The fix zero-initialises both result structs (struct midi_ump res = {0};), so the trailing words are cleared before transmission. These are the only two responder builders that left trailing words unset (send_string() already zeroes its buffer), so the leak is fully closed.

Netis NC63 firmware through V3.0.0.3327 contains a stack-based buffer overflow vulnerability that allows unauthenticated remote attackers to overwrite saved stack state by supplying an oversized destHost parameter to the ipFilterList=mod action in netis.cgi. Attackers can exploit widthless sscanf conversions that copy user-supplied input into fixed-size stack buffers before authentication is verified, achieving remote code execution as root due to the Boa web server executing the CGI environment with root privileges.

Netis NC63 firmware through V3.0.0.3327 contains a stack-based buffer overflow vulnerability that allows unauthenticated remote attackers to overwrite saved stack state by submitting an oversized Base64-encoded password to the login handler in /bin/netis.cgi. Attackers can exploit the custom Base64 decoder's lack of output length validation against the fixed-size stack buffer to achieve remote code execution with root privileges, as the Boa web server executes the CGI environment as root.

NetworkManager did not apply the private_user restriction to the 802-1x.ca-path and phase2-ca-path directory-valued connection properties. This incomplete fix for CVE-2025-9615 allows an unprivileged local user to point a private WPA-Enterprise (802.1X) connection profile's CA path at an attacker-controlled directory, bypassing server certificate validation and enabling credential theft via a rogue access point.

HCL Hive is affected by incorrect default permissions which could allow an attacker unauthorized lateral movement, container breakout, and interception of sensitive internal communications.

Craft CMS versions from 4.0.0-RC1 before 4.18.2 and from 5.0.0-RC1 before 5.10.6 contain an authenticated remote code execution vulnerability in control panel element-search condition handling. A JSON cleanse bypass in condition.config allows Yii behavior/event configuration keys to be interpreted after decoding, enabling command execution as the PHP/web user.

BookStack before 26.05.4 contains a broken access control vulnerability that allows authenticated API users with image-update or image-delete permissions to manipulate other users' avatars by exploiting missing content-type restrictions in the Image Gallery API endpoints. Attackers can supply a user avatar's ID to the API controller, which loads any image type without the web controller's gallery and drawio restrictions, and when the avatar's uploaded_to field matches a page ID accessible to the attacker, the authorization check passes allowing the attacker to rename, replace, or delete the target user's avatar without requiring user-management permission.

HCL Hive is affected by a use of vulnerable third-party components which could allow an attacker unauthorized access or compromise of the system by exploiting publicly documented security flaws.

A heap-based buffer overflow vulnerability exists in Konami's Metal Gear Online 3, originating from improper validation of lobby data fields related to kicked players. The affected function processes a list of kicked player identifiers using the lobby data key "kick_num" to determine the number of entries, and individual kicked player IDs supplied via keys in the format "kicked_id_%i". The function does not validate that "kick_num" falls within the expected bounds. The game design limits matches to a maximum of 16 players, and the corresponding buffer for storing kicked player IDs is sized accordingly. If "kick_num" exceeds this limit, the function continues writing the provided player IDs past the end of the intended buffer and into adjacent memory regions. These adjacent regions contain Steam callback handler structures responsible for processing lobby data updates, lobby messages, and other related events. By supplying an oversized "kick_num" value and appropriate "kicked_id_%i" fields, an attacker can overwrite fields within the callback handler structures, including function pointers and callback argument values. Successful exploitation may enable control-flow hijacking, potentially allowing arbitrary code execution within the game process.

A server-side request forgery (SSRF) vulnerability was found in multiple AWX notification backends. The webhook, Mattermost, Rocket.Chat, and Grafana notification backends use notification template URLs as direct HTTP request targets without validating the target address against private, loopback, or reserved IP ranges. An organization notification administrator can create notification templates pointing to internal or loopback addresses, causing the AWX control node to issue HTTP requests to services that are not externally accessible. Additionally, the webhook notification backend follows HTTP redirects and resends configured Basic Authentication credentials to redirect targets regardless of host change, allowing an attacker to exfiltrate notification credentials by redirecting to an attacker-controlled host. The Grafana backend sends its API key in the Authorization header to the configured target URL.

A path traversal vulnerability was found in AWX's project archive extraction. The project_archive action plugin extracts zip and tar archive members by joining the project directory path with the member filename without performing path normalization, boundary validation, or rejecting directory traversal sequences. A malicious archive containing members with path traversal components can write files to arbitrary locations on the execution node's filesystem outside the intended project directory. An attacker who controls the archive content, either through a compromised upstream source, a malicious archive URL, or a man-in-the-middle attack on a plain HTTP connection, can achieve arbitrary file writes as the user performing the extraction, potentially leading to remote code execution through mechanisms such as cron files, SSH authorized keys, or playbook content injection.

Horde IMP's AppleDouble MIME viewer writes an attacker-controlled attachment name into an HTML status block without escaping it. In lib/Mime/Viewer/Appledouble.php, _IMPrender() obtains the name of the data part with IMP_Contents::getPartName(), which returns the MIME part's own name parameter as supplied by the message, and passes it through sprintf into the text of an IMP_Mime_Status object. IMP_Mime_Status::__toString() concatenates each text entry directly into the surrounding table markup, so the value reaches the rendered page verbatim. A message crafted as multipart/appledouble whose data part carries markup in its name parameter therefore executes script in the context of any user who views it, and the payload persists in the mailbox. Exploitation requires no account on the target system, only the ability to send mail to a user. Version 7.2.0 escapes the value with htmlspecialchars(). The researcher additionally chains this flaw with the arbitrary file read of CVE-2026-58451, and reports that script running in an administrator's session can reach an application code-execution path.

Cross-site scripting in the Web Administration interface of Network Optix Nx Witness VMS before version 6.1.3 on Linux, Windows and MacOS allows an adjacent-network attacker to execute arbitrary JavaScript in the browser of an authenticated administrator and steal the administrator's session token, resulting in Administrator Account Takeover. An attacker who controls an Nx server on the same network segment can set that server's site name to a script payload, which executes when an administrator opens the "Merge with Another Site" dialog and the site selection list is displayed.Solution: Update to Nx Witness VMS version 6.1.3 or later.

The userspace syscall verifier z_vrfy_mbox_send() in drivers/mbox/mbox_handlers.c validated the nested msg->data/msg->size fields by reading them directly out of live userspace memory, and then forwarded the original, still-mutable userspace struct mbox_msg * pointer to z_impl_mbox_send() and the underlying driver. Between the access check and the driver's use of msg->data, the validated pointer could be replaced, leaving a time-of-check/time-of-use window. On a system built with CONFIG_USERSPACE, any unprivileged userspace thread may invoke the mbox_send() system call. A second thread sharing the caller's address space can race to overwrite msg->data with a supervisor (kernel) address after the verifier's bounds check has passed but before the driver dereferences it. The driver then reads from the attacker-chosen address in supervisor context (for example memcpy(&data32, msg->data, msg->size) in the NXP mailbox driver, whose bytes are subsequently emitted to the peer mailbox endpoint). The impact is a userspace-to-supervisor access-control bypass: disclosure of kernel memory contents (high confidentiality impact), or, for an invalid/unmapped target address, a faulting kernel read causing denial of service. The fix snapshots the entire struct mbox_msg into a kernel-stack copy with k_usermode_from_copy() and validates and forwards that immutable copy, closing the race.

Improper Neutralization of Special Elements used in an OS Command in the package manager component of Black Duck blackduck-c-cpp before 3.0.7 allows an actor able to create a file within the scanned build directory to execute operating system commands as the account running the scan. Filesystem paths encountered while traversing the scanned directory are interpolated into command strings that are executed through a shell without quoting or escaping, so shell metacharacters within those paths are interpreted rather than treated as literal text. No control over the build command or the tool's configuration is required.

TIM Flow before 26.0.6 contains an improper authorization vulnerability that allows any authenticated user to submit arbitrary SQL queries to a privileged dashboard Excel export endpoint intended for administrative use only. Attackers can craft and submit unauthorized SQL queries to the export endpoint to retrieve sensitive database contents as a downloadable spreadsheet, bypassing role-based access controls.

Invocation of Process Using Visible Sensitive Information in Black Duck blackduck-c-cpp 1.0.17 through 3.0.6 allows an actor able to execute code within the scanned project's build to obtain the Black Duck API token via the ambient process environment, which is inherited by subprocesses launched during build capture and signature scanning. This applies only where the token is supplied through the BLACKDUCK_API_TOKEN or BD_HUB_TOKEN environment variable. Upgrading does not remediate prior disclosure; any token supplied to an affected version through an environment variable should be rotated.

HCL Hive is affected by a missing rate limit which could allow an attacker unauthorized access via brute-force or credential stuffing attacks, or cause a denial of service.

TIM Flow before 26.0.6 contains a CRLF injection vulnerability that allows remote attackers to inject arbitrary HTTP headers and response body content by embedding unsanitized carriage return and line feed sequences in the rt URL parameter, which is reflected into Set-Cookie response headers. Attackers can craft malicious requests to induce authenticated users to execute arbitrary JavaScript in their browser context, enabling session token theft and account credential modification.

RansomLook contains a stored cross-site scripting (XSS) vulnerability in the cryptocurrency wallet detail view. Cryptocurrency addresses and blockchain names originating from external sources, including the public crowd-sourced ransomwhe.re feed, were stored without sufficient validation and later embedded directly into an inline JavaScript onclick handler. Although Jinja HTML autoescaping was applied, it does not provide adequate protection when untrusted data is inserted into a JavaScript string inside an HTML attribute. HTML entities such as ' are decoded by the browser's HTML parser before the resulting attribute is interpreted as JavaScript. Consequently, a specially crafted cryptocurrency address containing quote characters and JavaScript syntax could escape the intended string literal and execute arbitrary JavaScript when a user clicked the affected wallet's CSV export button. Because cryptocurrency information imported from an untrusted upstream could reach the vulnerable rendering path, exploitation may not require an authenticated RansomLook account if an attacker can introduce a malicious wallet record into a consumed external data source. Successful exploitation could allow attacker-controlled JavaScript to execute in the security context of the RansomLook web application, potentially exposing information accessible to the victim or performing actions with the victim's privileges. The patch mitigates the issue by validating cryptocurrency addresses and blockchain identifiers before storage, restricting them to a safe character set, and replacing the inline JavaScript handler with data-* attributes and an external event listener so wallet values are treated strictly as data rather than executable JavaScript.

RansomLook contains an authorization weakness in the web-based configuration editor exposed through the /admin/config endpoint. The endpoint requires an authenticated session but does not perform an explicit privilege or administrator authorization check before allowing access to configuration-management functionality. An authenticated low-privileged user able to access the endpoint can submit crafted configuration values that are written directly to the application's config/generic.json file. The affected functionality permits modification of configuration sections including notification, LDAP, SMTP, and general application settings. Successful exploitation could therefore allow an attacker to alter security-sensitive application behavior, redirect integrations or notifications, modify authentication-related configuration, disrupt external services, or render the RansomLook installation unavailable. The configuration editor also operated on a configuration file containing sensitive values such as passwords, tokens, secrets, and API keys. Although the affected version contains logic intended to prevent recognized secret values from being returned to the browser, exposing configuration management through insufficiently authorized web functionality significantly increases the impact of a compromised or low-privileged account. The patch resolves the issue by completely removing the /admin/config route and associated configuration-editing interface, preventing application configuration from being modified through the web UI.

phpIPAM before 1.8.2 contains an authentication bypass vulnerability in the REST API that allows unauthenticated attackers to gain full API access by exploiting an insecure object cache keying mechanism. The cache is keyed by lookup value alone without including the searched column, enabling an entry written during an app_id lookup to satisfy a subsequent app_code lookup, allowing attackers to use the numeric database row identifier as an API token to read, write, and delete all IP address management records.

RansomLook exposed sensitive operator-side scraping configuration through multiple unauthenticated API responses. Location records associated with ransomware groups and markets were returned largely verbatim to unauthenticated callers whenever the location itself was not marked as private. These records could contain internal fields such as header, which may include authentication headers, session cookies, or other credentials used to access monitored websites; init_script, which may contain logic used to bypass CAPTCHA, anti-bot protections, or paywalls; and browser, which discloses details about the scraping environment. An unauthenticated remote attacker could query the affected API endpoints and obtain these values. Leaked authentication material could potentially be replayed against the monitored service, while disclosure of scraping and bypass logic could allow site operators or other attackers to identify and defeat RansomLook's collection mechanisms. The patch introduces an explicit allowlist of fields permitted in public location records and strips all operator-side fields before returning data to unauthenticated users. The accompanying change from <string:postname> to <path:postname> appears to be a functional correction allowing legitimate post titles containing / and does not, based on this patch alone, represent the security issue.

An authenticated attacker with Editor access or alert.instances.external:write can submit an external Alertmanager alert containing a controlled generatorURL. The attacker is authorized to create the alert, but not to execute script in another user's Grafana session. Grafana renders alert.generatorURL directly as the Alert Details See source LinkButton href without URL-scheme sanitization or a safe-protocol allowlist. The click interceptor's :// heuristic can be bypassed by placing :// inside a JavaScript comment. When a user with read access clicks See source, the browser executes attacker-controlled JavaScript in the Grafana origin with the clicking user's permissions.

RansomLook contains insufficient resource validation in the analysis PDF generation functionality. Analysis documents are converted from Markdown to HTML and passed to WeasyPrint for PDF rendering. Prior to the fix, WeasyPrint used its default URL fetcher, allowing resource references contained in an analysis to be resolved without restrictions. An authenticated attacker able to create or modify an analysis could embed crafted resource references using schemes such as file:// or http://. When the analysis was subsequently rendered as PDF, WeasyPrint would process these references with the privileges and network access of the RansomLook server. A malicious file:// reference could cause the renderer to access arbitrary files readable by the RansomLook process, potentially exposing sensitive configuration, credentials, or other local data through rendered resources. Network URLs could cause the server to initiate requests to localhost, internal network services, or external systems, resulting in server-side request forgery (SSRF) and potentially bypassing network-level access restrictions. The patch introduces a dedicated WeasyPrint URL fetcher that permits only data: resources, the RansomLook report logo, and files contained within the analysis asset directory. Network resources and filesystem paths outside these explicitly permitted locations are rejected.

Multiple vulnerabilities on affected versions of Zscaler Client Connector allow remote code execution, giving an unauthenticated, unprivileged user the ability to execute arbitrary code in the ZCC context.

HCL Hive Keycloak IAM Instance is affected by insufficient granularity of access control which could allow an attacker unauthorized access to resources.

Multiple vulnerabilities on affected versions of Zscaler Client Connector allow local privilege escalation, giving an unprivileged user the ability to execute arbitrary code in a privileged context.

A locally exploitable buffer overflow bug can cause a local denial-of-service attack on affected versions of Zscaler Client Connector on Android and ChromeOS.

A remotely exploitable buffer overflow bug can cause a local and kernel denial-of-service attack on affected versions of Zscaler Client Connector on Windows.

An authentication bypass issue exists in communications between affected versions of the Zscaler Client Connector and the Zscaler Client Connector Portal.

RansomLook contains a path traversal vulnerability in the handling of the screen field associated with group posts. The GroupPost.get API handler concatenates the database-controlled screen value directly with the application's source/ directory and opens the resulting path without verifying that the resolved file remains within the intended directory. Because the screen field is free-form and can be populated either through the administrative post editor or through data imported from a remote RansomLook instance, a malicious upstream instance can provide traversal sequences such as ../config/generic.json. When the affected post is subsequently retrieved through the API, RansomLook resolves and reads the attacker-controlled path and returns the contents of the referenced file Base64-encoded in the API response. This can allow an attacker (being admin) controlling imported post data to read arbitrary files accessible to the RansomLook process, potentially exposing sensitive configuration data, API credentials, password hashes, or other application secrets. The attack does not require the malicious upstream to possess an account on the affected RansomLook instance. The vulnerability is addressed by resolving screen paths with os.path.realpath() and verifying that the resolved path remains beneath the application's source/ directory. Validation is performed both when values are written and immediately before files are read. Using canonical paths also prevents traversal through symbolic links that would bypass purely lexical path normalization checks.

RansomLook fails to enforce the privacy status of ransomware groups and markets when distributing newly collected victim posts to external notification channels. The post-processing logic checks whether an individual post is marked private but does not verify whether the group or market to which the post belongs is configured as private. As a result, newly parsed victim information associated with a private group or market may be automatically published through enabled Rocket.Chat, Mastodon, Bluesky, and e-mail notification channels despite the entity being explicitly configured to prevent public disclosure. A similar issue affects the public MISP feed. The feed previously determined privacy using groupinfo(), which only queries the group database. Consequently, victim information associated with private markets could be added to the public MISP feed because the corresponding market privacy flag was not evaluated. An attacker or other unauthorized party able to access these public notification channels or the MISP feed may obtain victim information that was intended to remain private. Depending on the collected data, this may disclose victim names, ransomware activity, incident information, or other information associated with privately monitored groups and markets. The fix introduces a common is_private_entity() check covering both groups and markets and prevents private entity posts from being distributed through external notification channels or the public MISP feed. Internal storage and dashboard alerting remain unaffected.

Ransomlook contains a Redis glob pattern injection vulnerability caused by insufficient neutralization of user-controlled input before it is incorporated into Redis SCAN MATCH patterns. The /api/health/<name> endpoint attempted to resolve the supplied name to a known group or market, but when resolution failed it fell back to using the attacker-controlled value directly in a Redis key pattern. An unauthenticated attacker could therefore supply Redis glob metacharacters such as *, ?, [ or ] to broaden the SCAN operation beyond the intended group. For example, requesting /api/health/* could enumerate health information, mirror slugs, and uptime series belonging to all groups and markets, including entities marked as private. Similar unsafe interpolation was present in /api/crypto/chain/<chain> and in the delete_manual_torrent() function. The latter represents a potentially destructive sink because a crafted infohash containing glob metacharacters could cause the scan to match torrent-health keys belonging to other torrents if attacker-controlled input can reach that function. The patch removes the unsafe fallback from the health endpoint and introduces glob escaping for user-controlled values before they are incorporated into Redis SCAN MATCH expressions.

HCL Hive is affected by a cryptographic primitive with a risky implementation which could allow an attacker unauthorized lateral compromise or widespread credential leakage if a single internal component is breached.

A vulnerability was identified in bytebot-ai bytebot 0.0.1. The affected element is an unknown function of the component Agent Execution Workflow. Such manipulation leads to infinite loop. The attack may be performed from remote. The exploit is publicly available and might be used. This vulnerability only affects products that are no longer supported by the maintainer.

This CVE ID has been rejected or withdrawn by its CVE Numbering Authority.

act starts an HTTP Artifacts V4 backend whenever a workflow uses actions/upload-artifact@v4 or actions/download-artifact@v4. The control-plane RPCs of that backend, including CreateArtifact, GetSignedArtifactURL, ListArtifacts, FinalizeArtifact and DeleteArtifact, accept a caller-supplied workflow_run_backend_id and never check that it belongs to the requester: validateRunIDV4 in pkg/artifacts/artifacts_v4.go parses the value and returns it with the comparison against the requesting task's run ID left commented out. The signed URLs the backend issues are authenticated by an HMAC whose key is hardcoded to the four bytes 0xba 0xdb 0xee 0xf0, identical in every build, computed over a concatenation of endpoint, expiry, artifact name and task ID with no length prefix or delimiter, so signatures are both forgeable and ambiguous between differing artifact name and task ID pairs. The --artifact-server-addr flag defaults to the host's outbound address rather than loopback, leaving the backend reachable from the surrounding network. Any client that can reach it may read, overwrite or delete the artifacts of a concurrently running job with no credentials, exposing build outputs such as secrets and deployment credentials and permitting their replacement before the owning job consumes them.

adm-zip 0.5.9 through 0.6.0 follows symbolic links at the extraction destination. Utils.sanitize in util/utils.js enforces containment by comparing only the string form of an archive entry name against the resolved extraction root, and Utils.writeFileTo opens the computed destination with fs.openSync(path, "w", 0o666), which resolves symbolic links and carries neither O_NOFOLLOW nor a pre-write fs.lstatSync check. When a path component at the destination already exists as a symbolic link pointing outside the extraction root, extractAllTo, extractAllToAsync and extractEntryTo write the entry contents through that link and then chmod its target, placing attacker-controlled content in a file outside the root without any traversal sequence appearing in the archive. Reaching the write requires overwrite to be enabled, because the preceding fs.existsSync check also resolves the link and otherwise declines. An attacker able to create a symbolic link inside a shared, reused or predictable extraction directory, such as a temporary directory or a continuous integration workspace, can overwrite any file the extracting process is permitted to write.

webpack-dev-middleware resolves a request to a local file in getFilenameFromUrl by testing the request pathname against a traversal guard and then slicing it at a fixed character offset. The guard, UP_PATH_REGEXP applied to path.normalize(`./${pathname}`), only matches ".." that stands as a whole path segment, while the containment test is the string comparison pathname.startsWith(publicPathPathname) and the file path is built as path.join(outputPath, pathname.slice(publicPathPathname.length)). When the configured publicPath has no trailing slash, a request such as GET /assets../.env against publicPath /assets yields the pathname /assets../.env, whose only dot-dot sits inside the segment "assets.." and so passes the guard, but the offset slice cuts within that segment and hands "../.env" to path.join, resolving one directory above outputPath. Reading a file from that path requires the middleware to be backed by the physical filesystem, which happens when writeToDisk is true or a custom outputFileSystem is supplied, since the default memfs volume holds only build output. Traversal depth is limited to a single directory because a separately delimited dot-dot segment is collapsed during URL parsing before the guard runs. The default publicPath value of "auto" resolves to "/" and is not affected. This is an incomplete fix for CVE-2024-29180: the guard and offset slice were introduced by that fix and are present in every release from 5.3.4, 6.1.2 and 7.1.0 onward.

The official Flair wheels for 0.15.0 and 0.15.1 still contain flair/models/clustering.py, whose ClusteringModel.load static method returns pickle.loads(joblib.load(str(model_file))) and so executes arbitrary Python while loading a model file. Loading a model supplied by an attacker therefore runs that attacker's code with the privileges of the loading process. This is the same sink and the same file as CVE-2024-10073, which records 0.15.0 as the fixed version on the basis that clustering support was dropped in that release; the module was removed from the documented API but remains present in the distributed artifact and reachable by importing flair.models.clustering directly, so the earlier record's fixed version does not hold for the shipped package.

The Mercado Pago Node.js SDK interpolates caller-supplied identifiers into API request paths without percent-encoding them, so characters that are structural in a URL survive into the outgoing request. The payment (get, capture, cancel), paymentRefund (create, total, list, get), advancedPayment (get, capture, cancel, update, updateReleaseDate) and disbursementRefund (create, createAll, listAll) clients build their path as a template literal, for example RestClient.fetch(`/v1/payments/${id}`, ...) in src/clients/payment/get/index.ts. A dot-dot or slash sequence in the identifier is normalised by the WHATWG URL parser and redirects the request to a different endpoint, and a question mark appends attacker-chosen query parameters, in both cases carrying the merchant's own access token. An application that forwards an identifier influenced by an untrusted party into one of these methods without an ownership check therefore allows that party to reach other resources within the merchant's token scope. The repository already contains the intended helper, encodePathParam in src/utils/path.ts, which pull request 451 applied to roughly 29 other clients while leaving these unchanged.

Xinference loads models with Hugging Face remote code execution unconditionally enabled, and before version 2.12.0 exposes no setting to disable it. Six loader call sites pass trust_remote_code=True as a literal or as an unconditional default: RerankModel._get_tokenizer in xinference/model/rerank/core.py, SentenceTransformerRerankModel.load in xinference/model/rerank/sentence_transformers/core.py, SentenceTransformerEmbeddingModel.load in xinference/model/embedding/sentence_transformers/core.py, FlagEmbeddingModel.load in xinference/model/embedding/flag/core.py, and two sites in xinference/model/llm/transformers/core.py where PytorchModel._sanitize_model_config and PytorchModel._get_components default the value to True. Because a caller with model launch access can register a model whose type is unknown and supply an arbitrary model path, the server reaches _auto_detect_type and then AutoTokenizer.from_pretrained, which imports and executes Python declared by the model directory's own tokenizer_config.json auto_map, running attacker-supplied code with the privileges of the worker process. Version 2.12.0 gates every site behind allow_trust_remote_code and the XINFERENCE_TRUST_REMOTE_CODE setting, permitting remote code only for bundled built-in models.

RustDesk's Windows clipboard redirection copies a peer-supplied length into a fixed-size caller buffer without an upper bound check. When an OLE paste consumer such as explorer.exe calls IStream::Read with a buffer of cb bytes, CliprdrStream_Read in libs/clipboard/src/windows/wf_cliprdr.c requests that many bytes of a remote file through cliprdr_send_request_filecontents and then executes CopyMemory(pv, clipboard->req_fdata, clipboard->req_fsize), where req_fsize is taken verbatim from the peer's CLIPRDR FileContentsResponse by wf_cliprdr_server_file_contents_response (req_fsize = fileContentsResponse->cbRequested) and is never clamped to cb anywhere in the chain. The function's only length comparison, req_fsize < cb, handles the short-read case and is evaluated after the copy has already occurred. A malicious or compromised peer that answers a small file-contents read with an oversized response therefore writes attacker-chosen data past the end of the paste consumer's heap buffer when the local user pastes clipboard file contents offered by the remote side. The file is a fork of FreeRDP's client/Windows/wf_cliprdr.c, where the same defect is CVE-2026-68579, fixed in FreeRDP 3.30.0.

RansomLook does not consistently enforce authorization checks when accessing groups, markets, and ransom notes marked as private. An unauthenticated or otherwise unauthorized remote attacker can access information associated with private entities through several web views and API endpoints. The affected functionality can disclose private group or market names, ransom-note content, and metadata associated with private groups. The /compare functionality can also be queried directly with the name of a private entity, allowing an unauthorized user to retrieve information such as post counts, mirror totals, and uptime even when the entity is excluded from the normal user interface. The patch explicitly adds a privacy check before returning this information. Ransom-note views, search results, and API endpoints were similarly missing consistent filtering. The fix introduces normalized private-group identifiers and alias handling, then rejects or filters notes associated with private groups before returning them to unauthorized callers. An attacker can exploit the issue remotely without authentication or user interaction, resulting in disclosure of information that was explicitly intended to be restricted to authorized users.