CVE-20XX-?
agent?

Just published.

The newest CVEs, straight from CVE.org and NVD.

410,277 CVEs · updated 2026-11-21 11:55 UTC

Reset all
  1. CVE-2026-117645

    Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting') vulnerability in phoenixframework phoenix_html allows an attacker who controls attribute names to inject attributes and script into rendered HTML via Phoenix.HTML.attributes_escape/1. attributes_escape/1 in lib/phoenix_html.ex escapes attribute values but not attribute names. key_escape/1 passes a binary key through html_escape, which neutralizes quotes and angle brackets but not whitespace, so a key such as x onmouseover ends the attribute and starts a new one. An atom key is only dasherized, so any character passes, including quotes and >. Nested data, aria and phx keys take the same paths. An application is affected when it builds attribute keys from untrusted input and renders them through HEEx <tag {@attrs}>, content_tag/3, or a direct call. Before 3.0.0 the same code lived in Phoenix.HTML.Tag. This issue affects phoenix_html: from 1.0.0 before 3.3.5 and from 4.0.0 before 4.4.0.

    XSS Web app Phoenixframework Phoenix Html

    5.3 MEDIUM 0.3% exploit odds Fix listed
  2. CVE-2026-119338

    Tesseract is an open source OCR engine. From 5.4.0 until 5.5.4, AppendLinePolygon in src/api/pagerenderer.cpp destroys the caller-owned right-to-left Pta accumulator while receiving it by value, leaving line_top_rtl_pts and line_bottom_rtl_pts dangling. On an upright Hebrew or Arabic page at the default PAGE XML level, POLYGONFLAG with a false skewed_flag enters the vulnerable calls, while LEVELFLAG greater than zero provides an alternative gate. The calls repeatedly read and free the stale object, causing a double free and use after free and allowing corrupted coordinates to reach RecalcPolygonline. This issue is fixed in version 5.5.4.

    Memory safety

    6.9 MEDIUM 0.1% exploit odds Fix listed
  3. CVE-2026-119339

    Tesseract is an open source OCR engine. Prior to 5.5.4, WeightMatrix::DeSerialize trusts a mode byte from a traineddata LSTM network and calls WeightMatrix::DeSerializeOld when kDoubleFlag is clear. In default FAST_FLOAT builds, DeSerializeOld invokes ASSERT_HOST instead of returning an error, so loading a crafted model through TessBaseAPI::Init terminates the process and causes denial of service. This issue is fixed in version 5.5.4.

    Denial of service

    5.5 MEDIUM 0.2% exploit odds Fix listed
  4. CVE-2026-117666

    Improper Validation of Certificate with Host Mismatch vulnerability in deadtrickster ssl_verify_fun allows an adversary in the middle to impersonate a TLS peer via a certificate that carries a partial-label wildcard. try_match_wildcard/4 in src/ssl_verify_hostname.erl anchors only the text before and after the * of an identifier such as f*.example.net, so the wildcard also consumes dots and the certificate is accepted for f.evil.example.net. RFC 6125 confines a wildcard to the left-most label. The same matcher lets *.168.1.1 match the IP literal 192.168.1.1. Whole-label wildcards are matched correctly. Erlang/OTP 20.0 and later check the hostname in ssl before a user verify_fun sees valid_peer, so a client that keeps server_name_indication set rejects the mismatch regardless. Exposed are clients that disable SNI, servers that verify client certificate hostnames with this library, and direct callers of verify_cert_hostname/2. This issue affects ssl_verify_fun: from 1.0.1 before 1.1.8.

    Crypto Library Deadtrickster Ssl Verify Fun

    9.1 CRITICAL 0.2% exploit odds Fix listed
  5. CVE-2026-78402

    Stack-based buffer overflow and out-of-bounds write vulnerability in Thales Sentinel Keys Server on Windows allows Overflow Buffers.  Sentinel Keys Server does not check the length of the first field in the input message. A local attacker may craft an input and send it to Sentinel Keys Server. It may lead to code execution as NT AUTHORITY\SYSTEM enabling local privilege escalation. This issue affects Sentinel Keys Server: versions lower than 1.3.8.0.

    Memory safety Server

    7.3 HIGH 0.1% exploit odds No fix listed
  6. CVE-2026-78403

    Stack-based buffer overflow and missing authorization vulnerability in Thales Sentinel Keys Server on Windows allows Overflow Buffers. Sentinel Keys Server does not check the length of the first field in the HTTP request. An attacker may craft an HTTP request and send it to Sentinel Keys Server. It may lead to remote code execution as NT AUTHORITY\SYSTEM. This issue affects Sentinel Keys Server: versions lower than 1.3.8.0.

    Memory safety Server

    9.3 CRITICAL 0.6% exploit odds No fix listed
  7. CVE-2026-62302

    MyTube is a self-hosted downloader and player for several video websites. Prior to version 1.10.3, `server.ts` calls `registerStaticRoutes` and `registerCloudRoutes` before `registerApiRoutes`, where `authMiddleware` and `roleBasedAuthMiddleware` are mounted. Because Express processes middleware in registration order, `staticRoutes.ts` and `cloudRoutes.ts` remain outside the authentication stack even when `loginEnabled=true`. Unauthenticated clients can access local videos through `/videos/*`, thumbnails through `/images/*` and `/images-small/*`, avatars through `/avatars/*`, subtitles through `/subtitles/*`, cloud thumbnail caches through `/api/cloud/thumbnail-cache/*`, and cloud signed URLs or redirects through `/cloud/videos/:filename` and `/cloud/images/:filename`. The frontend login wall does not protect direct media requests. Cloud routes return 404 when cloud storage is unconfigured, but that is not an authentication rejection; protected API routes correctly return 401. Combined with hidden-video path leaks, the flaw permits unauthenticated downloads of hidden media. Default loginEnabled=false single-user deployments have no authentication boundary to bypass. This issue is fixed in version 1.10.3.

    Auth & access Web app

    8.7 HIGH 0.5% exploit odds Fix listed
  8. CVE-2026-62333

    OpenVM is a performant and modular zkVM framework built for customization and extensibility. Before 1.7.0 in the 1.x line and in affected 2.0 and 2.1 prereleases, `MemoryMerkleAir` does not forbid Merkle expansion rows at `parent_height` = 0 or below while `expand_direction` != 0. A malicious prover can fabricate a subtree below a memory leaf and swap digests inside that subtree to change the committed memory root without a corresponding memory operation. The fraudulent rows still balance against the Merkle and compression buses and require no memory-bus operation for the changed leaf. Because the memory Merkle root binds VM memory state across continuation segment boundaries, this affects the soundness of OpenVM proofs using persistent memory. This issue is fixed in versions 1.7.0, 2.0.0-rc.3 and 2.1.0-preview.

    Other

    8.7 HIGH 0.4% exploit odds Fix listed
  9. CVE-2026-62334

    OpenVM is a performant and modular zkVM framework built for customization and extensibility. Before 1.7.0 in the 1.x line and in affected 2.0 and 2.1 prereleases, `OpenVmHalo2Verifier.verify` constructs raw proof calldata from user-controlled `proofData`, performs a `staticcall` to `address(this)`, and treats a successful call as proof acceptance. If the first four proof bytes match another external selector exposed by the wrapper, dispatch can invoke a successful non-verifier function instead of the generated Halo2 fallback verifier. Applications using the generated Solidity wrapper can consequently accept arbitrary app commitments and public values without checking the proof suffix, accumulator, app commitments or public values. This issue is fixed in versions 1.7.0, 2.0.0-rc.3 and 2.1.0-preview.

    Other

    8.7 HIGH 0.5% exploit odds Fix listed
  10. CVE-2026-62335

    OpenVM is a performant and modular zkVM framework built for customization and extensibility. Before 1.7.0 in the 1.x line and in affected 2.0 and 2.1 prereleases, `OpenVmHalo2Verifier.verify` accepts `appExeCommit` and `appVmCommit` as `bytes32` values without checking that either is less than the `BN254` scalar field modulus. Although public-value and proof-data lengths are checked, callers can provide non-canonical 256-bit encodings greater than or equal to that modulus at the Solidity wrapper boundary. Applications relying on the wrapper can therefore accept non-canonical app executable and app VM commitments, creating ambiguity between the supplied byte-level values and the field elements used by the verifier. This issue is fixed in versions 1.7.0, 2.0.0-rc.3 and 2.1.0-preview.

    Other

    6.9 MEDIUM 0.4% exploit odds Fix listed
  11. CVE-2026-62337

    SNARK Verifier is a generic (S)NARK verifier. Prior to 0.2.5, the Halo 2 aggregation circuit can mishandle proof points encoded as the point at infinity. `EcPointLoader::multi_scalar_multiplication` places a `Value::Assigned(_)` base paired with `Value::Constant(ONE)` into its unscaled variable-base list and passes it directly to `sum_with_const`. The default shim delegates to halo2-lib `EccChip::sum`, whose incomplete affine addition assumes nonidentity input points. A prover-controlled proof point encoded as `(0, 0)` can therefore enter the final aggregation sum through the scalar-one shortcut and affect the MSM or accumulator computation. Practical exploitability depends on the surrounding aggregation and final-decider checks; the advisory does not establish universal acceptance of invalid proofs. This issue is fixed in version 0.2.5.

    Other

    6.9 MEDIUM 0.3% exploit odds Fix listed
  12. CVE-2026-61694

    go-xrpl is a Go implementation of the XRP Ledger protocol. In versions through the June 18, 2026 revision identified by the advisory, `ParseMPTAmountBinary` in `internal/ledger/state/ripple_state.go` converts the eight-byte MPT magnitude using `num.Int64()` without checking that it fits in 63 bits. A positive magnitude of at least 2^63 silently wraps into a negative amount without an error. `ParseEscrow` passes 33-byte `sfAmount` fields to the parser and stores `MPTRaw()` in `EscrowData.MPTAmount`, causing incorrect downstream calculations and divergence from rippled's interpretation of the same escrow entry. Independently, `readMPTAmount` in `internal/tx/ledgerfields/decode_stream.go` accepts over-limit magnitudes and surfaces large decimal strings without enforcing the maximum signed 64-bit value. The public codec in codec/binarycodec/types/amount.go already rejects these values, leaving the internal decoders inconsistent with that safe path. A source fix is available in commit 9dbd3935ed6c54f0232f2291edf9c012a0900dd5.

    Memory safety Library

    3.7 LOW 0.3% exploit odds Fix listed
  13. CVE-2026-61850

    Kargo manages and automates the promotion of software artifacts. From 0.6.0 until 1.8.14, 1.9.9, and 1.10.8, two authorization flaws in the `project-scoped RBAC API` let an authenticated user with delegated permission to manage a Project's `Roles` or `ServiceAccounts` confer permissions exceeding the user's own. A user authorized to edit `Roles` can create, replace, or add rules granting permissions not held by the user; a user authorized to edit a Role's ServiceAccount can bind an identity, including the user's own, to a more powerful Role. A Kargo Role comprises a `ServiceAccount, Role, and RoleBinding` in the Project namespace. Changes are written using the API server's broadly privileged credentials after only a coarse resource-kind edit check, rather than checking each permission being conferred. This permits escalation to Project-administrator access, disclosure of `Project Secrets` and credentials, and modification or deletion of Project resources. Stolen Git, container-registry, or Helm-registry credentials can extend confidentiality and integrity impacts to those external systems. Escalation remains limited to Projects where the attacker already has RBAC-management delegation and does not itself grant cluster-wide or cross-Project access. Default non-administrator roles do not permit these operations; exploitation requires an `administrator-created custom role` granting the relevant permissions and control of an account assigned that role. This issue is fixed in versions 1.8.14, 1.9.9, and 1.10.8.

    Auth & access Server

    9.3 CRITICAL 0.4% exploit odds Fix listed
  14. CVE-2026-62284

    OSAC operator is a Kubernetes operator. In the affected revision identified as 8220502, the `console-proxy` `RemoteConfigResolver` parses Kubernetes Secret kubeconfig data with `clientcmd.RESTConfigFromKubeConfig` without sanitizing the resulting `rest.Config`. An attacker who can create or modify a Secret labeled `osac.openshift.io/remote-cluster-kubeconfig`: "true" in a namespace where the proxy's RoleBinding grants Secret read access can supply `users[].user.exec`. When `rest.TransportFor` builds the HTTP transport for an upstream WebSocket connection, client-go invokes that exec credential plugin, spawning an arbitrary process as the osac-`console-proxy` service account inside the pod. Attacker-controlled `clusters[].server` also enables server-side request forgery from the proxy pod. Remote and auto VM cluster modes are affected, with auto the default; exclusively `local` mode does not read kubeconfig Secrets and is unaffected. The credential-plugin execution path is patched in commit 7e4222b; that patch does not address the described server-address SSRF path.

    Injection Web app

    8.6 HIGH 0.5% exploit odds Fix listed
  15. CVE-2026-62301

    MyTube is a self-hosted downloader and player for several video websites. Prior to version 1.10.3, backend video access does not enforce the visitor role's intended public-only restriction when `loginEnabled=true` and a visitor password is configured. The frontend filter in `VideoContext.tsx` is only client-side, `roleBasedAuthMiddleware.ts` allows all visitor GET requests, and `storageService/videos.ts` performs `getVideos()` and `getVideoById(id)` queries without checking the caller's role or `visibility`. A visitor can enumerate `/api/videos` or request `/api/videos/:id` to retrieve hidden `visibility`=0 records, exposing titles, `sourceUrl` values, `videoPath` values, subtitle paths and `thumbnailPath` values. The `serveMountVideo()` path also serves mounted video files without `visibility` checks. Leaked paths work against unauthenticated `/videos/*` and `/subtitles/*` routes, allowing media downloads without credentials. The RSS visitor feed already filters `visibility`=1 content and is not the affected query path. Default loginEnabled=false single-user deployments are unaffected by this role-boundary issue because all callers are owner-equivalent. This issue is fixed in version 1.10.3.

    Auth & access Web app

    7.1 HIGH 0.4% exploit odds Fix listed
  16. CVE-2026-55598

    PinePods is a complete, self-hosted podcast management system written in Rust. Prior to 0.9.0, an authenticated user can subscribe to an attacker-controlled RSS feed through `POST /api/data/add_podcast` in `rust-api/src/handlers/podcasts.rs`. `get_podcast_values` and `add_episodes` in `rust-api/src/database.rs` parse the feed with `feed-rs` and store the `<enclosure url>` verbatim in `Episodes.episodeurl` without destination host/IP validation. `POST /api/data/download_podcast` invokes `spawn_download_podcast_episode` and `download_episode_and_wait()` in `rust-api/src/services/tasks.rs`, which sends a `reqwest` GET to the stored URL and saves and records the response in `DownloadedEpisodes` for later retrieval. The `estimate_duration_from_audio_url_async()` path in `rust-api/src/database.rs` also sends an unvalidated HEAD during feed add/refresh. Default redirect following permits an initially external enclosure URL to redirect to an internal target. A low-privilege user can make server-side GET/HEAD requests to reachable loopback, private-network and cloud metadata services and read downloaded internal responses, enabling internal-service enumeration, confidential-data disclosure and possible metadata credential theft. Episode downloading requires server downloads to be enabled, which is the default; self-service signup is supported. This issue is fixed in version 0.9.0.

    Disclosure Server

    9.1 CRITICAL 0.3% exploit odds Fix listed
  17. CVE-2026-61618

    SNARK Verifier is a generic (S)NARK verifier. Prior to 0.2.5, KZG accumulator public-instance limbs are decoded without consistently enforcing canonical limb bounds before elliptic-curve coordinate reconstruction. `PlonkProof::read` selects scalars using `accumulator_indices` and passes them to `AccumulatorEncoding::from_repr`. For the EVM loader, `LimbsEncoding::from_repr` forwards them to `EvmLoader::ec_point_from_limbs` in `snark-verifier/src/loader/evm/loader.rs`, where `shl` and `add` reconstruct coordinates before `validate_ec_point` checks the resulting affine point rather than the original limb bounds. The related native path in `snark-verifier/src/pcs/kzg/accumulator.rs` can reach assertion- or unwrap-based decoding and panic instead of returning a verification error. This allows noncanonical encodings at the verifier boundary and can crash the native verifier. Standard SDK aggregation circuits are expected to constrain accumulator limbs before this boundary, which limits practical exploitability. This issue is fixed in version 0.2.5.

    Denial of service

    6.9 MEDIUM 0.4% exploit odds Fix listed
  18. CVE-2026-61620

    HoneySQL turns Clojure data structures into SQL. From 2.0.0 until 2.7.1392, applications that pass user input or other arbitrary data into the HoneySQL `DSL` by calling `keyword` or `symbol` on that input can produce SQL strings that facilitate SQL injection when executed unchanged by a `JDBC` driver. Some formatters do not apply the existing suspicious-character check to the resulting entity names. The impact depends on the privileges of the database connection executing the generated statement. Input validation before constructing `keyword` or `symbol` entities is a workaround. The maintainer-designated fixed version is 2.7.1392.

    Injection OS & kernel

    9.8 CRITICAL 0.4% exploit odds Fix listed
  19. CVE-2026-55098

    Arc is an open, SQL-native time-series database for telemetry. Prior to the repository-verified 26.06.2 repair, the remediation of `CVE-2026-47735` leaves a cross-database authorization bypass through DuckDB table-valued I/O functions. In `internal/database/sandbox.go`, `allowed_directories` uses `cfg.LocalStorageRoot`, configured by `storage.local_path`, while `enable_external_access=false` blocks access outside that entire multi-tenant root. `GetStoragePath` in `internal/storage/util.go` stores all databases beneath it. In `internal/api/query.go`, `extractTableReferences` skips names followed by an opening parenthesis and `checkQueryPermissions` allows queries with no extracted table references, including tokens with `permissions: []`. The `userSQLReadParquetPattern` denylist catches literal `read_parquet(` and the earlier protection also blocks `arc_partition_agg(`, but misses the `parquet_scan` alias and `glob`, `parquet_metadata`, `parquet_schema`, `read_blob`, `read_csv_auto`, `read_json` and `read_text`. A request to `POST /api/v1/query` using these unblocked functions passes `ValidateSQLRequest` and `convertSQLToStoragePaths` without tenant authorization, allowing reads and enumeration of other databases' time-series data, schema and per-column value-range statistics within the allowlisted root. The sandbox still blocks out-of-root OS files and HTTP SSRF, and the sibling SQLite token database outside the root is not exposed by this mechanism. The earlier 26.06.1 remediation does not comprehensively deny these I/O functions. Repository tag 26.06.2 contains the I/O-function repair and supplies the provisional remediation boundary; conflicting advisory affected-version statements require curator adjudication.

    Disclosure Server

    6.5 MEDIUM 0.3% exploit odds Fix listed
  20. CVE-2026-55493

    HotCRP is conference review software. Starting with 2.101, the `cmt:` search term checks whether a comment is visible to the searching user but fails to check whether the commenter's identity should be visible. A `cmt:USER` search can therefore match a paper with a visible comment authored by USER even when the comment is supposed to be anonymous. An authenticated user with access to the search can infer the author of an anonymous comment, compromising reviewer anonymity. Comments that the user cannot see remain excluded from search results. The advisory prose identifies 3.3 as fixed, its affected-version metadata excludes 3.3, and the verified repository tag 3.3 already contains the identity-check fix; however, its patched-version metadata instead lists 3.3.1. This fixed-version source conflict remains unresolved.

    Auth & access

    6.5 MEDIUM 0.3% exploit odds Fix listed
  21. CVE-2026-119335

    Pi-hole is a DNS sinkhole that protects devices from unwanted content without installing any client-side software. Prior to 6.5, advanced/Scripts/utils.sh getFTLPID() attempts to validate the pihole-owned /run/pihole-FTL.pid file with expr "${FTL_PID}" : "[^[:digit:]]", but this rejects only values whose first character is not a digit. An attacker with code execution as the pihole user can write a digit-prefixed arithmetic expression to the pihole-owned PID file. Root-run pihole subcommands then evaluate [[ "$pid" -eq "-1" ]], causing Bash to recursively evaluate array subscripts and execute command substitution as root. This bypasses the content-validation portion of the CVE-2026-41489 remediation while leaving its hardcoded PID-file path protection intact. This issue is fixed in version 6.5.

    Injection

    7.8 HIGH 0.2% exploit odds Fix listed
  22. CVE-2026-119336

    Pi-hole is a DNS sinkhole that protects devices from unwanted content without installing any client-side software. From 6.4.2 until 6.5, the root-run advanced/Scripts/updatecheck.sh cron job applies truncate -s 0 and chmod 644 to /etc/pihole/versions while following symbolic links in a directory owned by the pihole user. An attacker executing as pihole can replace that path with a symlink so the scheduled root operations truncate or create an arbitrary root-owned file, or race the operations so chmod makes a sensitive target world-readable without truncating it. This issue is fixed in version 6.5.

    Other

    7.8 HIGH 0.2% exploit odds Fix listed
  23. CVE-2026-119337

    RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine. From 0.25.2 until 0.27.3, POST /api/v1/agents/chat/completions invokes completion_openai and returns through the openai-compatible branch before UserCanvasService.accessible() checks whether the caller may access the requested agent. An authenticated user who knows another user's private agent UUID can execute that Only me workflow and read its output across the user or tenant boundary; configured tools may also cause side effects, but stronger impacts were not demonstrated. This issue is fixed in version 0.27.3.

    Auth & access Web app

    7.1 HIGH 0.3% exploit odds Fix listed
  24. CVE-2026-119730

    Maxthon Browser for Android before 7.5.8.103 contains an address bar spoofing vulnerability that allows remote attackers to misrepresent a page's origin by exploiting the lack of a persistent fullscreen notification or trusted origin indicator when the Fullscreen API is invoked. Attackers can craft a malicious web page that renders a counterfeit address bar showing a trusted domain while displaying attacker-controlled content, tricking users who perform a single tap to enter fullscreen into disclosing credentials via phishing.

    Other Browser & mail

    5.1 MEDIUM 0.2% exploit odds No fix listed
  25. CVE-2026-119331

    Ladybird is a truly independent web browser. Prior to the c0d68bddf3e3616e5173835f0f4c155ee9700d91 commit, CSSStyleSheet.cssRules, insertRule(), and deleteRule() in Libraries/LibWeb/CSS/CSSStyleSheet.cpp did not enforce the stylesheet origin-clean flag with the required SecurityError, even though cross-origin imported stylesheets were correctly marked non-origin-clean. A malicious page could read parsed rules from a stylesheet loaded without CORS and could insert or delete rules in the victim stylesheet, bypassing the Same-Origin Policy. Reading could expose selectors, generated content, feature markers, or other sensitive rule text, while mutation could alter page visibility or layout. The issue does not provide memory corruption, code execution, or persistence after the attacker page closes. This vulnerability is fixed by commit c0d68bddf3e3616e5173835f0f4c155ee9700d91.

    Disclosure Browser & mail

    7.1 HIGH 0.2% exploit odds Fix listed
  26. CVE-2026-119332

    Ladybird is a truly independent web browser. Prior to 209579c96473c75bd5eba9701fcadb841a045964, FinalizationRegistry::cleanup() in Libraries/LibJS/Runtime/FinalizationRegistry.cpp removed a dead record from m_records, retained an iterator to the next FinalizationRecord, and then invoked attacker-controlled JavaScript through the cleanup callback. If that callback called unregister() for the next record, remove_by_token() freed the node referenced by the retained iterator, and the next cleanup iteration read its target and held_value fields after free. This reliably crashes instrumented WebContent builds and creates a memory-corruption primitive in production builds. This vulnerability is fixed by commit 209579c96473c75bd5eba9701fcadb841a045964.

    Memory safety Browser & mail

    8.8 HIGH 0.2% exploit odds Fix listed
  27. CVE-2026-119333

    draw.io Desktop is the official Electron build of draw.io. Prior to 31.5.2, the Windows application launched powershell and attrib by bare executable name from src/main/electron.js. When draw.io Desktop was not already running, opening a diagram from File Explorer made the diagram directory the current working directory, allowing a colocated attacker-controlled powershell.bat or attrib.exe to be resolved before the intended system executable. A victim who extracted an attacker-supplied archive and cold-opened its legitimate .drawio file could therefore execute attacker-controlled code when Local Fonts were enabled or when draft or backup attributes were updated. This issue is fixed in version 31.5.2.

    Code execution

    7.8 HIGH 0.1% exploit odds Fix listed
  28. CVE-2026-119334

    Pi-hole is a DNS sinkhole that protects devices from unwanted content without installing any client-side software. Prior to 6.5, gravity.sh checks file:// adlist readability with sudo -u pihole test -r but then allows curl, sometimes running as root, to normalize URL dot segments and open a different file. A pihole-readable decoy can therefore pass the check while curl reads a root-only target, and /proc/self paths can similarly resolve against different processes. Exploitation requires an administrator who can add adlists and can place a symlink and decoy on the filesystem. This issue is fixed in version 6.5.

    Disclosure

    2.2 LOW 0.3% exploit odds Fix listed
  29. CVE-2026-119328

    Immich is a high-performance self-hosted photo and video management solution. Prior to 3.3.0, the unauthenticated machine-learning /predict endpoint in machine-learning/immich_ml/main.py accepted an options object and passed its values into model construction in machine-learning/immich_ml/models/base.py without restricting security-sensitive parameters. An attacker with network access to the service could set cache_dir to an arbitrary writable path, causing model downloads to write there and causing clear_cache() to pass that path to shutil.rmtree() after selected model-loading failures. Depending on filesystem permissions and the selected path, this could modify files in mounted data locations or recursively delete an attacker-selected directory. The model-name allowlist prevents this issue from becoming arbitrary-model deserialization or code execution. This issue is fixed in version 3.3.0.

    Disclosure Web app

    6.5 MEDIUM 0.4% exploit odds Fix listed
  30. CVE-2026-119329

    Immich is a high-performance self-hosted photo and video management solution. Prior to 3.0.0, GET /api/map/markers used server/src/services/map.service.ts to add partner owner identifiers when withPartners=true, while server/src/repositories/map.repository.ts applied Archive visibility to every selected owner when isArchived=true. A view-only partner could therefore enumerate the GPS coordinates and existence of a sharer's archived photos or videos even though the timeline and direct-access paths exclude archived partner assets. The exposure is limited to location metadata and asset identifiers because image bytes remain protected by checkPartnerAccess. This issue is fixed in version 3.0.0.

    Auth & access Web app

    4.3 MEDIUM 0.3% exploit odds Fix listed
  31. CVE-2026-119330

    Immich is a high-performance self-hosted photo and video management solution. Prior to 3.2.0, the shared-link creation API accepted albumId and assetIds as independent fields in server/src/services/shared-link.service.ts but authorized only the resource selected by the requested link type. An authenticated attacker who already knew a victim-owned asset UUID could select SharedLinkType.Album for an attacker-owned album while supplying that UUID in assetIds; SharedLinkService.create() persisted the unchecked association in shared_link_asset and made the victim asset available through the new anonymous shared-link credential. This exposed victim metadata, thumbnails, and the byte-identical original file and created a persistent sharing relationship that could be redistributed. This issue is fixed in version 3.2.0.

    Other Web app

    3.5 LOW 0.2% exploit odds Fix listed
  32. CVE-2026-116249

    In server mode pgAdmin 4 resolves a server's password file, SSL certificate, key, CRL and SSH tunnel identity file relative to the user's storage directory, but get_complete_file_path() only stripped leading slashes before joining the two. A path such as "../<another user>/.pgpass", set through the server dialog, the REST API or a servers.json import, therefore named another user's file, and connecting used that user's saved passwords, client certificate or SSH key. The sslkeylogfile connection parameter was passed to libpq unchanged, so a path outside storage let libpq create or append a file as the account running pgAdmin. make_conninfo() also quoted connection parameter values but not their names, so a crafted parameter name could smuggle additional keyword=value pairs past these checks. The fix requires the resolved path, with symbolic links followed, to stay inside the user's storage directory or the named shared storage directory, confines sslkeylogfile the same way, and accepts only exact libpq keywords as connection parameter names. A refused path or keyword is reported instead of letting libpq fall back to a default. This issue affects pgAdmin 4: before 9.19.

    Disclosure Server Pgadmin 4

    7.2 HIGH 0.2% exploit odds Fix listed
  33. CVE-2026-116250

    In server mode pgAdmin 4 validates a path a user gives for a file in their storage directory with check_access_permission() and then opened it by name. Several places opened it with a plain open() or handed the path to a utility that opened it later: the servers.json export and import and the import file listing, and the files given to pg_dump, pg_dumpall, pg_restore and psql by Backup, Restore and Import/Export Data. A symbolic link swapped in at the file or at any parent directory between the check and the open redirected the operation outside the storage directory, so a file could be written, overwritten or read, or loaded into a table, as the operating-system account running pgAdmin. A hard link to an outside file, which O_NOFOLLOW does not detect, and a FIFO, which could hang a request, had the same effect. CVE-2026-86861 fixed this for the File Manager's save_file; these were the remaining paths. Exploitation requires creating such a link or FIFO inside the user's storage directory. pgAdmin offers no way to do so, and the per-user storage directories belong to the pgAdmin account, so this requires write access to the storage backend as a different operating-system user, for example a shared or network-mounted storage directory, and winning the race. The fix opens each file by walking down from the storage directory one component at a time with O_NOFOLLOW, accepting only a regular file with one link, and passes the utilities the open descriptor (/dev/fd/N) on Linux. Error messages no longer include the server path. This issue affects pgAdmin 4: before 9.19.

    Disclosure Server Pgadmin 4

    7.3 HIGH 0.1% exploit odds Fix listed
  34. CVE-2026-116251

    pgAdmin 4 did not invalidate a user's existing sessions when their password changed, either through the user's own change password dialog or through an administrator resetting it in User Management. A session established before the change, for example on another device or by someone who had obtained the old password or a session cookie, therefore stayed logged in after the change, which is the step a user or administrator takes precisely to cut such access off. It remained usable until it expired. The fix rotates the user's session identifier (fs_uniquifier) whenever the password is changed, so every other session of that user is logged out. The session that made the change is logged in again and keeps its database connections. Saved server passwords, including those for shared servers, are re-encrypted under the new key; any that cannot be decrypted are cleared and the user is told. If re-encryption fails, the sessions are still revoked and the saved passwords are cleared rather than the change failing. This issue affects pgAdmin 4: before 9.19.

    Other Web app Pgadmin 4

    2.3 LOW 0.1% exploit odds Fix listed
  35. CVE-2026-116252

    pgAdmin 4's materialized view refresh passed the target database's name to psql's --dbname argument unchanged. libpq treats a --dbname value containing "=" (or a postgresql:// URI) as a full connection string, whose keywords override the host, port and other settings pgAdmin had supplied. A database whose name is such a string, which any role with CREATEDB can create, therefore redirected the psql process pgAdmin starts for the refresh to a host and port of the creator's choosing, and the connecting user's saved server password was presented to that endpoint. It also let the pgAdmin host be made to open connections to arbitrary addresses. The same weakness was fixed for Backup, Restore, Maintenance and Import/Export Data in earlier releases (CVE-2026-86862); the materialized view refresh was the remaining caller. The fix routes the name through the same database_conninfo() helper, so libpq treats it strictly as a database name. This issue affects pgAdmin 4: before 9.19.

    Other Server Pgadmin 4

    7.1 HIGH 0.2% exploit odds Fix listed
  36. CVE-2026-116243

    pgAdmin 4 runs pg_dump, pg_restore, pg_dumpall and psql from the binary path a user sets in their preferences. In server mode the binary path was refused only when it lay under the per-user storage directory (STORAGE_DIR), not under a SHARED_STORAGE directory, which ordinary users can upload files to. A user could therefore upload an executable to shared storage, point their binary path at it, and have pgAdmin run it as the operating-system account running pgAdmin. On POSIX, files created through pgAdmin are not executable, so this is exploitable where any file with an executable name can run, such as Windows server mode. The fix checks the binary path against every SHARED_STORAGE directory as well, after expanding $DIR, collapsing ".." and resolving symbolic links, compares directories in a way a different spelling on a case-insensitive filesystem cannot bypass, refuses a path containing a NUL, and enforces the check wherever binary paths are resolved, including for the PSQL tool. This issue affects pgAdmin 4: before 9.19.

    Code execution OS & kernel Pgadmin 4

    7.7 HIGH 0.3% exploit odds Fix listed
  37. CVE-2026-116244

    When a pgAdmin 4 connection was not meant to use a Kerberos user's delegated credentials, pgAdmin removed KRB5CCNAME from its environment, so the Kerberos library fell back to the default credential cache, or a client keytab, of the operating-system account running pgAdmin. libpq then answered any server that asked for GSSAPI authentication with that account's ticket, whatever user name the pgAdmin user gave, so in server mode any user could connect as the service account's Kerberos principal; gssencmode=disable did not prevent it. On Windows, SSPI used the service account's logon in the same way. psql was also started with pgAdmin's environment, which could hold another user's credential cache while that user was connecting. The fix points KRB5CCNAME and KRB5_CLIENT_KTNAME at a file that cannot exist except while a Kerberos-authenticated user connects with Kerberos, and gives every other connection, psql and the utilities require_auth=!gss,!sspi and gssencmode=disable, checked on the final connection string. require_auth needs libpq 16 or later. This issue affects pgAdmin 4: before 9.19.

    Auth & access OS & kernel Pgadmin 4

    7.1 HIGH 0.3% exploit odds Fix listed
  38. CVE-2026-116245

    libpq looks a connection service up in the service file of the operating-system account running pgAdmin 4, and an entry there can hold a password or name a password file or client certificate. In server mode any user could set a server's Service, or a "service" connection parameter, to one of those entries and connect with its credentials. A PGPASSWORD or PGSERVICE variable set for that account was likewise used by libpq for every connection without a value of its own and passed on to the utilities, and pg_dumpall read a maintenance database name such as "service=..." as a connection string. The fix adds the SERVER_MODE_ALLOWED_SERVICES setting, empty by default: in server mode only a listed service can be used, the servicefile parameter is refused, and PGPASSWORD and PGSERVICE are removed from pgAdmin's environment at startup. The allowlist is checked again on the final connection string, and the pg_dumpall database name is wrapped as a connection string. This issue affects pgAdmin 4: before 9.19.

    Auth & access Server Pgadmin 4

    7.1 HIGH 0.2% exploit odds Fix listed
  39. CVE-2026-116246

    To pass a server's password to psql and to the utilities pgAdmin 4 runs as background processes, the server manager wrote it into pgAdmin's own process environment: as PGPASSWORD for the PSQL tool, and under a per-process variable for a utility. Nothing removed it afterwards. pgAdmin's environment is shared by every user's requests and copied into every process it starts, so in server mode another user's later PSQL session or utility inherited the password, and libpq used PGPASSWORD for any connection made without a password of its own, letting one user connect with another user's saved password. The fix puts the password only into the environment built for the process that needs it and starts psql with a copy of pgAdmin's environment instead of the environment itself. This issue affects pgAdmin 4: before 9.19.

    Other Server Pgadmin 4

    7.0 HIGH 0.2% exploit odds Fix listed
  40. CVE-2026-116247

    When a server in pgAdmin 4 had no password file, client certificate or client key configured, or the configured one could not be found, pgAdmin left the parameter out of the connection and libpq fell back to the defaults of the operating-system account running pgAdmin: ~/.pgpass, ~/.postgresql/postgresql.crt and postgresql.key, or PGPASSFILE, PGSSLCERT and PGSSLKEY. In server mode any user could therefore connect to a server with the service account's saved passwords or client certificate, and supplying only a copy of its public certificate was enough for libpq to pair it with the service account's private key. A pg_service.conf entry, or a maintenance database name read as a connection string by pg_dumpall, could bring the same files back for the utilities. The fix gives each of these parameters a value in server mode, the user's own file or otherwise a path that can never be opened, in the connection string, in the environment of the utilities and in the --dbname connection string that ranks above a service entry. Desktop mode is unchanged. This issue affects pgAdmin 4: before 9.19.

    Auth & access Server Pgadmin 4

    7.1 HIGH 0.1% exploit odds Fix listed
  41. CVE-2026-116248

    In pgAdmin 4 server mode only administrators are meant to share a server with all users; the server dialog shows the Shared switch only to them. The create and update endpoints of the server API, however, stored the "shared" flag and shared username they were sent, so any user able to register a server could share one with every other user. The servers.json import had a similar gap: a non-administrator's "Shared" entries were removed only after validation, so a file that also held an invalid entry could create a shared server. A shared server carries no credentials, so this gives no direct access. It lets a user plant a server, named to look trusted, in every other user's tree and pointing at a host the user controls; anyone who then connects to it presents their own database password to that host. The fix refuses a non-administrator's request that would share a server with HTTP 403, prunes a non-administrator's shared entries from a servers.json import before validating it, and warns at startup about servers shared earlier by non-administrators. This issue affects pgAdmin 4: before 9.19.

    Auth & access Network gear Pgadmin 4

    6.0 MEDIUM 0.2% exploit odds Fix listed
  42. CVE-2026-116239

    In server mode pgAdmin 4 keeps an account per authentication source, so the same username can exist as, for example, an internal account and an LDAP account. After LDAP or Kerberos authenticated a user, pgAdmin chose the account to log in by username alone, or took the account Flask-Security's login form had looked up, without checking that it came from the source that authenticated the user. The session therefore belonged to another account, with its role, its servers and its saved database passwords: - An LDAP or Kerberos user whose username matched an internal account was logged in as that internal account, including the administrator account created when pgAdmin is first set up. - With internal authentication listed before LDAP or Kerberos in AUTHENTICATION_SOURCES, the login form's internal validation looks up the account whose email was typed and keeps it even though its password does not match. A Kerberos user could type another Kerberos user's principal and be logged in as that user, and an LDAP user could do the same with an LDAP account whose email is their username. Exploitation requires valid credentials for an LDAP directory or Kerberos realm that pgAdmin is configured to use. The fix looks the account up by the authenticated username and the source that authenticated it, and no longer uses the account the login form found for any source but internal authentication. Internal authentication logs in that account only if it is an internal account. Desktop mode is not affected. This issue affects pgAdmin 4: before 9.19.

    Auth & access Server Pgadmin 4

    8.9 HIGH 0.4% exploit odds Fix listed
  43. CVE-2026-116240

    In server mode pgAdmin 4 validates a path into the user's storage directory with check_access_permission(), which builds the absolute path with get_abs_path(). Several callers then acted on a path they built differently, so the path checked and the path used could name different files: - Opening a file in the Query Tool, the Google client secret and, on Windows, Import/Export Data, Backup, Restore and saving a file joined the path to the storage directory after stripping leading slashes and backslashes. On POSIX a path checked as a name inside storage could become an absolute path once stripped, so any file the pgAdmin account can read could be opened; on Windows a path naming a drive discarded the storage directory, so these could read, or write, outside it. - Listing a folder decoded the path a second time after the check, so a folder named like an encoded "..", which a user can create, listed any directory on the host. - Download, delete, rename and the file-exists check appended the path to the storage directory without a separator, so in a deployment where one user's storage directory name is a prefix of another's, another user's files could be read, deleted or moved; download also checked the path appended to itself. - Creating a folder could create one in another user's directory. The fix makes check_access_permission() return the path it checked and every caller act on that path, decodes folder paths once, and opens, deletes and renames files relative to directories opened from the storage directory without following symbolic links. Desktop mode is not affected. This issue affects pgAdmin 4: before 9.19.

    Disclosure Server Pgadmin 4

    7.1 HIGH 0.3% exploit odds Fix listed
  44. CVE-2026-116241

    pgAdmin 4's Import/Export Data tool wraps a user's query export in a psql \copy (...) TO command and checks that the query's parentheses are balanced so that nothing can escape the wrapper. The check rejected a backslash inside a single-quoted string, but not inside a double-quoted identifier, where it treated the next double quote as closing the identifier. psql's \copy parser applies the connection's escape character to whichever quote opened a token, so with standard_conforming_strings off a backslash-escaped double quote does not close the identifier. pgAdmin's and psql's view of the quoting then diverged, a closing parenthesis hidden from the check ended the \copy (...) wrapper, and a following TO PROGRAM clause was executed by psql on the pgAdmin host, as the operating-system account running pgAdmin, before the server rejected the statement. An ordinary database role can turn standard_conforming_strings off for itself, and the Import/Export Data permission is granted to non-administrators by default. This is a bypass of the fix for CVE-2026-17566. The fix rejects any backslash inside a double-quoted identifier, as was already done for single-quoted strings, since its meaning depends on a server setting that cannot be known reliably when the query is validated. This issue affects pgAdmin 4: before 9.19.

    Injection Server Pgadmin 4

    9.4 CRITICAL 0.3% exploit odds Fix listed
  45. CVE-2026-116242

    pgAdmin 4 opens SSH tunnels with the sshtunnel library but gave it neither host_pkey_directories nor ssh_config_file, so it always loaded the unencrypted private keys in ~/.ssh of the operating-system account running pgAdmin, read that account's ~/.ssh/config (IdentityFile, User, HostName, ProxyCommand) for the tunnel host, and could use its ssh-agent. In server mode any user could therefore configure a tunnel to a host that accepts the service account's key, with no credentials of their own or with an identity file that did not exist in their storage, and reach it as that account. The fix makes tunnels in server mode use only the identity file or password the user configured: the service account's keys, configuration and agent are no longer used, and an identity file that cannot be found in the user's storage is reported instead of falling back to ~/.ssh. Desktop mode is unchanged. This issue affects pgAdmin 4: before 9.19.

    Other Library Pgadmin 4

    7.1 HIGH 0.2% exploit odds Fix listed
  46. CVE-2026-115829

    In Secret Server on-prem versions earlier than 12.3.1, a user could cause a two-factor pin code to be generated for any account without prior authentication, then bypass the initial login step and authenticate using the second factor alone.

    Auth & access Server

    9.9 CRITICAL 0.5% exploit odds No fix listed
  47. CVE-2026-115830

    In Secret Server on-prem versions earlier than 12.3.1, a user under certain conditions could submit an unsigned JWT and receive cloudadmin permissions in response.

    Auth & access Server

    8.7 HIGH 0.3% exploit odds No fix listed
  48. CVE-2026-116238

    In server mode pgAdmin 4 locks an internal account after MAX_LOGIN_ATTEMPTS failed logins (3 by default). The limit could be bypassed, giving an unauthenticated attacker an unlimited number of password guesses against any internal account, including the administrator's, in the default configuration: - Flask-Security registers its own views next to pgAdmin's. A login submitted to its /login view was checked and logged in without pgAdmin's failed login count or lock. Its /verify view checked a signed-in user's password without limit, its /change view changed a password without pgAdmin's change password permission or re-encrypting the user's saved server passwords, and its /reset views sent and applied password resets for accounts that are not internal or are locked. - The internal account was looked up by its exact username, while the login form matches the email ignoring case, so a login with the username typed in another case was neither counted nor checked for the lock. - With LDAP or webserver authentication also configured, a wrong internal password was never counted, and with Kerberos or webserver authentication configured a guess could be answered without being counted. In addition, where accounts from different authentication sources share an email address, a password reset could be sent for, or applied to, an account from another source, and the About box showed the server configuration to a user whose email was that of an administrator's account. The fix refuses submissions to Flask-Security's own views, counts every failed internal login however the username is typed and whatever other sources are configured, no longer lets a login that posts a password fall back to an authentication source that does not check it, and locks only internal authentication. Password resets and the About box act only on the user's own internal account. Desktop mode is not affected. This issue affects pgAdmin 4: before 9.19.

    Auth & access Server Pgadmin 4

    9.2 CRITICAL 0.4% exploit odds Fix listed
  49. CVE-2026-115828

    In Secret Server on-prem versions earlier than 12.3.1, Secret Server did not always validate the PIN or biometric check when using a FIDO2-compliant device

    Auth & access Server

    9.4 CRITICAL 0.3% exploit odds No fix listed
  50. CVE-2026-94175

    Unauthenticated Cross Site Scripting (XSS) in All In One Login <= 2.4.1 versions.

    XSS Web app

    7.1 HIGH 0.2% exploit odds No fix listed

The newest published CVE records from CVE.org and NVD, refreshed every few minutes, with severity, exploit odds and fix status.