CVE Database

sysPass's FileBackupService::doBackupFiles() in lib/SP/Services/Backup/FileBackupService.php around line 388 builds a tar shell command by string-concatenating the backup directory path $this->path directly into the command line ('tar czf ' . $backupFileApp . ' ' . BASE_PATH . ' --exclude \"' . $this->path . '\" 2>&1') and passes the result to PHP's exec() with no application of escapeshellarg() and no validation of the path against a safe character set. The $this->path value is read from the sysPass configuration, which is persisted in the database and writable through the admin settings API and the admin UI. An administrator (or an attacker who has obtained an admin API token or admin session) can therefore store a backup path containing shell metacharacters and trigger a backup operation to execute arbitrary OS commands as the web server process user (typically www-data or apache). Because sysPass is a password manager whose sole purpose is to hold credentials for other systems, code execution as the web-server user permits reading sysPass's master password and encryption key from memory or configuration files, decrypting every stored credential in the database, exporting the entire password vault, pivoting to internal systems using the disclosed credentials, and installing persistent backdoors on the password-manager host.

Dgraph is an open source distributed GraphQL database. Prior to 25.3.8, maybeQuoteArg in graphql/resolve/query_rewriter.go passes regexp filter strings into generated DQL without quoting or validating the /pattern/flags form, allowing crafted GraphQL query or mutation filters to inject DQL operators, disclose unintended nodes, or expand modification and deletion targets. This issue is fixed in version 25.3.8.

Fedify is a TypeScript library for building federated server apps powered by ActivityPub. From version 1.2.0 through the affected 1.9, 1.10, 2.0, 2.1, 2.2, and 2.3 maintenance lines, getNodeInfo() follows an attacker-controlled links[].href value from /.well-known/nodeinfo without scheme, redirect, or private-address validation, allowing requests to loopback, link-local, cloud metadata, and private-network services and returning their response bodies. This issue is fixed in versions 1.9.13, 1.10.12, 2.0.22, 2.1.18, 2.2.7, and 2.3.2.

Contiki-NG's MQTT client parse_publish_vhdr() in os/net/app-layer/mqtt/mqtt.c sets topic_len_received=1 before checking topic_len against the 64-byte limit, so an over-length topic returns early but leaves the flag set. On the next TCP segment, tcp_input() re-invokes the parser with topic_received==0, and the persisted topic_len_received==1 skips the length-reading block containing the guard, falling through directly to a memcpy() that uses the unvalidated 16-bit topic_len as the copy length. The 65-byte topic[] destination overruns into adjacent struct fields including the payload_chunk pointer, which subsequent MQTT code dereferences, giving a compromised or attacker-controlled broker an arbitrary-pointer-write primitive. Contiki-NG's MQTT implementation has no TLS support so the connection is plaintext. Impact ranges from information disclosure and denial of service to remote code execution on embedded targets without memory protection.

Contiki-NG's DNS/mDNS resolver skip_name() in os/services/resolv/resolv.c walks DNS wire-format name labels with no packet-boundary check, and the caller in newdata() invokes it in a loop iterating nquestions times from the attacker-controlled DNS header before validating the transaction ID. An attacker who sets nquestions higher than the number of complete questions present causes skip_name() to walk past the UDP packet buffer, and the returned pointer is cast to struct dns_answer * for further memory reads. On builds with RESOLV_CONF_SUPPORTS_MDNS enabled, any peer on the local segment can trigger the read unauthenticated via a multicast UDP 5353 packet with no outstanding query required; on standard DNS builds an attacker who can inject a UDP response from port 53 during an outstanding query can trigger the same read. Impact is out-of-bounds read of uip_buf and adjacent memory, disclosing memory contents or crashing the resolver.

Contiki-NG's LwM2M TLV parser lwm2m_tlv_read() in os/services/lwm2m/lwm2m-tlv.c ignores its caller-supplied buffer length argument and reads up to six bytes from the input buffer with no bounds check. The caller in lwm2m-engine.c iterates while there is at least one byte remaining, so a crafted CoAP WRITE to any LwM2M endpoint whose final TLV supplies exactly one byte triggers up to five out-of-bounds reads of heap memory adjacent to the CoAP input buffer, disclosing memory contents (including key material and peer addresses) through the parsed tlv->id, tlv->length, and tlv->value fields. Corrupted tlv_len derived from the out-of-bounds memory further corrupts the caller's parse offset. In LwM2M NoSec mode, the default for constrained devices, no authentication is required.

Ground Station prior to 0.6.0 contains an unauthenticated blind server-side request forgery vulnerability in the orbital-source configuration path that allows any unauthenticated Socket.IO client to cause the ground-station process to issue outbound HTTP requests to attacker-chosen destinations. Attackers can connect to the Socket.IO server on port 7000 without credentials due to disabled authentication enforcement and a wildcard CORS policy, then submit a data_submission event with submit-orbital-sources action to persist an attacker-supplied URL in the database, then trigger an orbital sync via the equally unauthenticated background_task:start event. The URL is stored with no scheme allowlist, no host validation, and no rejection of loopback, RFC1918, or link-local (cloud instance metadata at 169.254.169.254) addresses, and is passed directly to requests.get in _fetch_http_3le and _fetch_http_omm in backend/tlesync/source_adapters.py. HTTP status codes and error messages from the outbound request are emitted in the orbital_sync_state Socket.IO event to all connected clients, providing a serviceable oracle for interpreting internal-service and cloud-metadata responses even though the raw response body is not directly leaked. Because the malicious source persists in the database across restarts and re-fires every 24 hours on the scheduled sync cycle, the primitive gives durable long-term SSRF without the attacker needing to remain connected.

OpenReception's appointment booking software provides an end-to-end encrypted appointment booking platform. Versions prior to 1.0.2 don't throttle failed passphrase login attempts. An attacker can submit unlimited wrong passphrase guesses against any known email address, capped only by the Argon2 verification cost (about 100 milliseconds per attempt on the tested host, giving 10 attempts per second sustained). The same backend implements a working per-account throttle on the WebAuthn challenge endpoint, which returns HTTP 429 after roughly 19 attempts. The passphrase branch simply does not invoke that throttle, leaving a supported high-value login path unprotected against credential stuffing and dictionary attacks. The asymmetry confirms this is an oversight rather than a design choice. The throttle infrastructure exists, is wired into the same auth backend, and works on the WebAuthn path. The passphrase branch in `/api/auth/login` was not updated to record failed attempts. Combined with the application's minimum-passphrase policy (12 characters, no entropy or dictionary checks), accounts using common base patterns such as `Spring2026!XX` or words from a leak corpus are realistically reachable in days on a single CPU, hours on a small GPU farm. Version 1.0.2 patches the issue.

OpenReception's appointment booking software provides an end-to-end encrypted appointment booking platform. Prior to version 1.0.2, a TENANT_ADMIN can store `javascript:` URLs in the tenant `links` configuration (`website`, `imprint`, `privacyStatement`). These values are returned to the patient-facing landing page via `/api/public`, hydrated into the SvelteKit Button component, and rendered as `<a href="javascript:...">` elements without URL-scheme filtering. A patient who clicks any such link executes the attacker's JavaScript inside the patient browser origin, where patient form data is read before client-side encryption is applied. This breaks the project's central trust claim that the server is an untrusted relay and that administrators cannot read patient data. Patient-side encryption happens after form input, so JavaScript executing in the patient origin can read or alter the plaintext before encryption is performed. Version 1.0.2 fixes the issue.

OpenReception's appointment booking software provides an end-to-end encrypted appointment booking platform. Prior to version 1.0.2, the `GET /api/tenants/{id}` endpoint returns the full tenant record to any authenticated `TENANT_ADMIN` of that tenant, including the `databaseUrl` field. This field contains the live PostgreSQL connection string the application uses to connect to that tenant's database. In the tested official `docker-compose.prod.yml` deployment, the connection string contained the user `postgres` with `rolsuper=true` and the plaintext password from `secrets/postgres_password.txt`. Operators who configure a non-superuser PostgreSQL user via `secrets/postgres_user.txt` would expose a less privileged credential, but the disclosure of the connection string itself is independent of that choice. The same credential applies to every database managed by that PostgreSQL instance: the central `appointment_booking` database, every per-tenant database (one per tenant), and the postgres administrative database. A `TENANT_ADMIN` of one tenant who can reach `postgres:5432` (directly via internal network, indirectly via any SSRF, RCE, or file-read in the application) can read every other tenant's appointment ciphertexts, key shares, and metadata; read the central user table, including all `GLOBAL_ADMIN` accounts, password hashes, and session records; modify or delete any data in any tenant database; and/or i a superuser-scoped deployment: use PostgreSQL's `pg_read_server_files`, `COPY ... FROM PROGRAM`, and `CREATE EXTENSION` for further escalation inside the database container. This breaks the per-tenant database isolation that is otherwise the primary cross-tenant control in the application. The application code carefully scopes most queries to the calling tenant's database, but those scopings are irrelevant once the attacker holds the credentials that bypass the application entirely. Version 1.0.2 fixes the issue.

OpenReception's appointment booking software provides an end-to-end encrypted appointment booking platform. Prior to version 1.0.2, when a user navigates to the `/logout` page, the page's server-side load handler deletes the `access_token` cookie before calling `/api/auth/logout` via an internal `event.fetch()`. The internal fetch consequently runs without the auth cookie, so `apiAuthHandle` rejects it, the logout handler never executes, and `SessionService.revokeSession()` is never called for the current session. The DB session row remains valid until its natural expiry (one week by default). The user sees a successful logout (cookie gone, UI returns to login), but any party still holding a copy of the now-deleted access token can continue making authenticated API calls until the session naturally expires. The root cause is a simple ordering mistake. The same auth subsystem implements the correct order in `/api/auth/logout`: revoke the current DB session first, then delete the cookie. The page-level wrapper does the opposite. Version 1.0.2 initiates server-side logout before removing authentication cookies and first appears in version 1.0.2. Version 2.0.0 later replaces this with a race-free client-side logout flow.

OpenZeppelin Contracts Wizardis a web application to interactively build a contract out of components from OpenZeppelin Contracts. Versions prior to 0.10.9 generate a Hardhat test file (`test/test.ts`) by interpolating user-supplied `opts.name` (ERC20/ERC721) and `opts.uri` (ERC1155) directly into TypeScript string literals at `zip-hardhat.ts:48` and `:50` without any JavaScript string escaping. No authentication is required: an attacker crafts a URL such as `https[:]//wizard[.]openzeppelin[.]com/#/erc20?name=");require("child_process").execSync("...");("` and shares it with a developer. When the victim downloads the resulting zip archive and runs `npx hardhat test`, the injected Node.js code executes with the developer's local OS privileges. Version 0.10.9 fixes the issue.

Frappe is a full-stack web application framework. Prior to 15.110.0 and 16.20.0, the restore and bulk_restore endpoints do not apply the appropriate document permission checks, allowing an authenticated user to restore deleted documents without the required authorization. This issue is fixed in versions 15.110.0 and 16.20.0.

Frappe is a full-stack web application framework. Prior to 15.108.0 and 16.18.3, temporary magic login link generation can use an attacker-controlled request Host header, allowing a remote attacker to cause emailed login links to point to an attacker-controlled domain and capture the login token when a recipient follows the link. This issue is fixed in versions 15.108.0 and 16.18.3.

Decidim is a participatory democracy framework. Prior to 0.31.5 and in 0.32.0.rc1 before 0.32.0.rc2, JWT-backed API authentication is not bound to the organization selected by the current host, allowing a JWT issued for one tenant to be replayed against another tenant’s API to read participantDetails data and reach the proposal.answer mutation path. This issue is fixed in versions 0.31.5 and 0.32.0.rc2.

Decidim is a participatory democracy framework. Prior to 0.30.9, from 0.31.0 before 0.31.5, and in 0.32.0.rc1 before 0.32.0.rc2, the identity-document verification admin UI embeds verification_attachment blobs through reusable signed Active Storage disk URLs, allowing anyone who obtains a URL to download the scanned document without an authenticated Decidim session until the signature expires. Verification-document images are rendered with variant_url(...), which produces signed /rails/active_storage/disk/... links instead of routing the file through an authorization-checking controller. Because Decidim configures Active Storage service URLs to remain valid for seven days, the URL itself becomes the credential for that period. The affected files are verification_attachment blobs on Decidim::Authorization, and the admin review pages embed those signed URLs directly into the HTML for pending and confirmation views. This issue is fixed in versions 0.30.9, 0.31.5, and 0.32.0.rc2.

llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in llama-server affecting six tokenization endpoints (/tokenize, /detokenize, /infill, /apply-template, /rerank, and /anthropic/count_tokens) that bypass the task queue and access ctx_server.vocab directly on HTTP worker threads. Attackers can exploit a time-of-check-time-of-use race condition where the main thread destroys and frees vocab after the synchronization lock is released but before the handler finishes using it, causing a crash or potential code execution when --sleep-idle-seconds is configured.

llama.cpp builds b7492 through the latest b9060 contains a use-after-free vulnerability in the vocab pointer of llama-server when the --sleep-idle-seconds feature is enabled, allowing unauthenticated remote attackers to execute arbitrary code. Attackers can trigger the vulnerability by sending requests to affected endpoints while the server transitions to sleep mode, causing concurrent worker threads to dereference a freed vocab pointer that can be reclaimed with attacker-controlled data to achieve remote code execution.

llama.cpp builds b4882 through b9058 contain a heap buffer overflow vulnerability in the KV cache state restore path where the state_read_data() function computes write size without overflow checking, allowing attackers with write access to the slot_save_path directory to corrupt heap memory. Attackers can craft malicious state files where cell_count multiplication overflows or exceeds tensor buffer allocation to write attacker-controlled bytes past buffer boundaries, potentially resulting in heap metadata corruption, model weight corruption, or arbitrary code execution via function pointer overwrite.

llama.cpp builds b3978 through b9058 contain an integer underflow and out-of-bounds read vulnerability in the DRY sampler that allows unauthenticated attackers to trigger a heap buffer underflow by sending a crafted HTTP request with dry_allowed_length set to INT32_MIN to the /v1/completions or /v1/chat/completions endpoints. Attackers can exploit this vulnerability to crash the server with SIGSEGV causing denial of service for all connected users, or corrupt token sampling probabilities by reading garbage values from memory before the allocated buffer.

llama.cpp builds b1283 through b9058 contain an integer overflow vulnerability in the llama_batch_init() function where unchecked multiplications in malloc() calls can wrap past INT32_MAX when computing allocation sizes. Attackers can pass specially crafted parameters to trigger integer overflow, causing heap corruption and potentially achieving arbitrary code execution through subsequent batch operations that write past allocated buffer boundaries.

The XML and schema validation functionalities within the SchemaValidator Mediator process XML input as part of validation flows. Under certain conditions, the XML parser allows the resolution of external entities when handling user-supplied XML content during validation operations. This behavior can occur when an attacker supplies crafted XML payloads to the relevant mediator flows with sufficient privileges. Successful exploitation may allow a highly privileged actor to read files accessible within the server hosting the affected product. Additionally, it may be possible to trigger outbound requests to unintended internal or external locations, depending on the server environment and network configuration. Specially crafted XML payloads can also lead to excessive resource consumption during parsing, impacting the availability of the product.

A maliciously crafted PDF file, when parsed through Autodesk Revit, can force a Use-After-Free vulnerability. A malicious actor can leverage this vulnerability to cause a crash, disclose sensitive data, or execute arbitrary code in the context of the current process.

Insufficient validation of untrusted input in UI in Google Chrome prior to 151.0.7922.109 allowed a remote attacker who had compromised the renderer process to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High)

Use after free in Skia in Google Chrome prior to 151.0.7922.109 allowed a remote attacker who had compromised the renderer process to execute arbitrary code inside a sandbox via a crafted HTML page. (Chromium security severity: High)

Integer overflow in V8 in Google Chrome prior to 151.0.7922.109 allowed a remote attacker to execute arbitrary code inside a sandbox via a crafted HTML page. (Chromium security severity: High)

Out of bounds write in Skia in Google Chrome prior to 151.0.7922.109 allowed a remote attacker who had compromised the renderer process to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: High)

Use after free in Views in Google Chrome prior to 151.0.7922.109 allowed a remote attacker who had compromised the renderer process to potentially perform a sandbox escape via a crafted HTML page. (Chromium security severity: Critical)

Insufficient validation of untrusted input in Contextual Tasks in Google Chrome prior to 151.0.7922.109 allowed a remote attacker to perform privilege escalation via a crafted HTML page. (Chromium security severity: High)

Inappropriate implementation in V8 in Google Chrome prior to 151.0.7922.109 allowed a remote attacker to execute arbitrary code inside a sandbox via a crafted HTML page. (Chromium security severity: High)