Search: server.py
1,271 threats
Most recently published first
| Severity | CVSS | Signals | Threat | Source | Published | Score and star |
|---|---|---|---|---|---|---|
| Low | 2.7 | EPSS 0.5% | CVE-2026-50269aio-libs aiohttp · CRLF Injection Low2.7EPSS 0.5%CVE Database V5CVE-2026-50269 AIOHTTP is an asynchronous HTTP client/server framework for asyncio and Python. Prior to 3.14.0, attacker-controlled input included into multipart/payload headers can be used to modify a request to inject additional headers or similar. In the unlikely situation that an application is passing user-controlled strings into MultipartWriter.append(headers=...) or Payload.headers, then an attacker may be able to modify the request to inject headers or change the contents of the request. This vulnerability is fixed in 3.14.0. | CVE Database V5 | ||
| High | 7.2 | EPSS 0.9% | CVE-2026-7246SUSE click · Security update for python-click High7.2EPSS 0.9%GCVE DatabaseCVE-2026-7246 Pallets Click versions 8.3.2 and below contain a command injection vulnerability in the click.edit() function. This flaw allows attackers to execute arbitrary operating system commands from an unprivileged account. A security update addressing this issue is available. The vulnerability has been rated with medium severity. There are no known exploits in the wild at this time. | GCVE Database | ||
| Medium | 6.9 | EPSS 0.5% | CVE-2026-11748LY Central Dogma · LDAP Injection Medium6.9EPSS 0.5%CVE Database V5CVE-2026-11748 # Vulnerability `SearchFirstActiveDirectoryRealm.findUserDn()` substitutes the user-supplied username from the login form into an LDAP search filter template (default `cn={0}`) **without escaping RFC 4515 filter metacharacters** (`*`, `(`, `)`, `\`, NUL). Combined with `SearchControls.setCountLimit(1)` on the same call site, this allows three distinct attack primitives: 1. **Authentication confusion** — typing username `*` causes the realm to construct filter `cn=*`, return the first directory entry (typically a privileged account in AD ordering), and attempt bind against that DN with the attacker's password. 2. **Audit log evasion** — payload `bob)(uid=alice` is recorded verbatim in audit logs while the realm searches with the malformed filter, breaking accountability/compliance (SOX, PCI-DSS, ISO 27001). 3. **Directory enumeration** — wildcards and timing differences allow reconnaissance of OU structure and admin group membership. A repo-wide search for any LDAP escape helper (`escapeLdap`, `encodeFilter`, `escapeFilter`, `ldapEscape`) returns **zero hits** — the defense is not just missing, it was never added. > **Applicability note:** This realm is opt-in. The shipped default LDAP example (`dist/src/conf/shiro.example.ldap.ini`) uses Shiro's `DefaultLdapRealm` with `userDnTemplate` and is **NOT** affected. However, the realm exists precisely to support Active Directory environments where users log in via `sAMAccountName` and the realm must search for the DN first — the canonical LINE corporate AD-backed SSO scenario. Internal deployments using AD-backed login almost certainly select this realm. --- ## Evidence **File:** `server-auth/shiro/src/main/java/com/linecorp/centraldogma/server/auth/shiro/realm/SearchFirstActiveDirectoryRealm.java` **Lines 148–176** on branch `main` @ commit `d64a5151`: ```java @Nullable protected String findUserDn(LdapContextFactory ldapContextFactory, String username) throws NamingException { LdapContext ctx = null; try { ctx = ldapContextFactory.getSystemLdapContext(); final SearchControls ctrl = new SearchControls(); ctrl.setCountLimit(1); // line 156 — returns FIRST match only ctrl.setSearchScope(SearchControls.SUBTREE_SCOPE); ctrl.setTimeLimit(searchTimeoutMillis); final String filter = searchFilter != null ? USERNAME_PLACEHOLDER.matcher(searchFilter) .replaceAll(username) // line 162 — RAW SUBSTITUTION : username; // line 163 final NamingEnumeration result = ctx.search(searchBase, filter, ctrl); ... ``` `USERNAME_PLACEHOLDER = Pattern.compile("\\{0}")`. Default `searchFilter = "cn={0}"`. ### Data flow from HTTP login to vulnerable substitution | Step | Component | |------|-----------| | HTTP login form | `POST /api/v1/login` form field `username` | | `ShiroLoginService.usernamePassword()` (lines 198–223) | applies `loginNameNormalizer` (Unicode lowercase only — **NOT** LDAP escape) | | `Subject.login(new UsernamePasswordToken(username, password))` | Shiro hand-off | | `ActiveDirectoryRealm.doGetAuthenticationInfo` (Shiro core) | calls `queryForAuthenticationInfo0` | | `SearchFirstActiveDirectoryRealm.findUserDn(factory, upToken.getUsername())` | username flows in **verbatim** | ### Repository-wide escape helper grep | Search term | Hits | |-------------|------| | `escapeLdap` | 0 | | `encodeFilter` | 0 | | `escapeFilter` | 0 | | `ldapEscape` | 0 | --- ## PoC Self-contained JUnit 5 test using UnboundID `InMemoryDirectoryServer` (in-process, no external LDAP required). Drop into `server-auth/shiro/src/test/java/com/linecorp/centraldogma/server/auth/shiro/realm/LdapInjectionPoCTest.java` and add `com.unboundid:unboundid-ldapsdk:7.0.0` as a test dependency. > The PoC works by subclassing the realm and overriding `findUserDn()` to capture the actual LDAP filter string sent to the directory — the captured filter is the structural evidence, independent of LDAP server strictness about bind outcomes. ```java /* * Copyright 2026 LINE Corporation * * SECURITY PoC — NOT FOR MERGE INTO THE MAIN TEST SUITE. * * This JUnit class demonstrates the LDAP filter injection in * SearchFirstActiveDirectoryRealm. Drop into * server-auth/shiro/src/test/java/com/linecorp/centraldogma/server/auth/shiro/realm/ * Adds the UnboundID LDAP SDK as a test dep. */ package com.linecorp.centraldogma.server.auth.shiro.realm; import static org.assertj.core.api.Assertions.assertThat; import static org.assertj.core.api.Assertions.assertThatThrownBy; import javax.naming.directory.SearchControls; import javax.naming.ldap.LdapContext; import org.apache.shiro.realm.ldap.JndiLdapContextFactory; import org.apache.shiro.realm.ldap.LdapContextFactory; import org.junit.jupiter.api.AfterAll; import org.junit.jupiter.api.BeforeAll; | CVE Database V5 | ||
| Critical | 9.4 | EPSS 0.2% | CVE-2026-11746LY Central Dogma · CWE-798 Critical9.4EPSS 0.2%CVE Database V5CVE-2026-11746 ## Vulnerability `ZooKeeperReplicationConfig.secret()` silently substitutes the hard-coded constant `"ch4n63m3"` (leetspeak for "change me") whenever the operator omits `replication.secret`. The same secret is wired into both the **client-facing SASL context** and the **quorum/learner SASL contexts** of the embedded ZooKeeper. The constant is in OSS source on GitHub and is discoverable via code search in seconds. ### Three Reinforcing Defects 1. **OSS-public credential** — `DEFAULT_SECRET` is in `line/centraldogma` source. 2. **Silent fallback** — `firstNonNull(convertValue(...), DEFAULT_SECRET)` substitutes the default with no log, no warning, no startup banner. The only sanity check `checkArgument(!secret().isEmpty(), ...)` passes because the getter substitutes the literal before the emptiness check runs. 3. **Dual-purpose secret** — used for both ZK client-port super auth and inter-peer quorum SASL. A single leaked password authenticates against both surfaces. ### Architecture Context (Important) Central Dogma does **NOT** connect to an external ZooKeeper ensemble. Each replica embeds a `QuorumPeer` (`EmbeddedZooKeeper extends QuorumPeer`) inside its own JVM. The Central Dogma cluster **IS** the ZK ensemble. So the "ZK network" is the inter-replica network of the Central Dogma cluster itself. ### Applicability | `replication.method` | ZK Started? | Applicable? | |---|---|---| | `NONE` (standalone, dev default) | No | **NOT applicable** | | `ZOOKEEPER` (HA production) | Yes, embedded on every replica | **Fully applicable** — canonical production configuration | --- ## Evidence **File:** `server/src/main/java/com/linecorp/centraldogma/server/ZooKeeperReplicationConfig.java` **Branch:** `main` @ commit `d64a5151` **Line 53** — the constant: ```java private static final String DEFAULT_SECRET = "ch4n63m3"; ``` **Lines 210–215** — the silent fallback: ```java /** * Returns the secret string used for authenticating the ZooKeeper peers. */ public String secret() { return firstNonNull(convertValue(secret, "replication.secret"), DEFAULT_SECRET); } ``` --- **File:** `server/src/main/java/com/linecorp/centraldogma/server/internal/replication/ZooKeeperCommandExecutor.java` **Lines 586–607** — JAAS wiring (same secret on both surfaces): ```java final String escapedSecret = jaasValueEscaper.escape(cfg.secret()); ImmutableList.of("Server", EmbeddedZooKeeper.SASL_SERVER_LOGIN_CONTEXT).forEach(name -> { buf.append(name).append(" {").append(newline); buf.append(DigestLoginModule.class.getName()).append(" required").append(newline); buf.append("user_super=\"").append(escapedSecret).append("\";").append(newline); buf.append("};").append(newline); }); ImmutableList.of("Client", EmbeddedZooKeeper.SASL_LEARNER_LOGIN_CONTEXT).forEach(name -> { buf.append(name).append(" {").append(newline); buf.append(DigestLoginModule.class.getName()).append(" required").append(newline); buf.append("username=\"super\"").append(newline); buf.append("password=\"").append(escapedSecret).append("\";").append(newline); buf.append("};").append(newline); }); ``` --- **File:** `server/src/main/java/com/linecorp/centraldogma/server/internal/replication/EmbeddedZooKeeper.java` **Line 44** — proves CD embeds the ZK server: ```java final class EmbeddedZooKeeper extends QuorumPeer { ``` **Lines 213–220** — client port binding (loopback only): ```java private static ServerCnxnFactory createCnxnFactory(QuorumPeerConfig zkCfg) throws IOException { final InetSocketAddress bindAddr = zkCfg.getClientPortAddress(); final ServerCnxnFactory cnxnFactory = ServerCnxnFactory.createFactory(); // Listen only on 127.0.0.1 because we do not want to expose ZooKeeper to others. cnxnFactory.configure(new InetSocketAddress("127.0.0.1", bindAddr != null ? bindAddr.getPort() : 0), zkCfg.getMaxClientCnxns()); return cnxnFactory; } ``` > Quorum/election ports are **NOT** loopback-bound — they bind to `replication.servers[].host` as configured, exposed on the inter-replica network. --- ## PoC Two attack surfaces, two scenarios. **Surface A** (client port, same-host) is implemented as a working read-only PoC. **Surface B** (quorum-port peer impersonation) is documented but intentionally not weaponized. ### Surface A — Same-Host Client Port (Loopback) PoC Python + kazoo + pure-sasl. Authenticates as `super` over SASL DIGEST-MD5 with the leaked secret and reads the full Central Dogma replication log. Hardcoded to `127.0.0.1`, read-only, prints first 5 entries. ```python #!/usr/bin/env python3 """ C3 PoC -- ZooKeeper default-secret takeover (read-only, loopback only). Demonstrates that a Central Dogma instance launched with a ZooKeeper-replicated configuration but without `replication.secret` set exposes its embedded ZooKeeper to anyone with local-host access, using the well-known credential `super / ch4n63m3`. SAFETY: * Hardcoded to 127.0.0.1. Refuses any other targ | CVE Database V5 | ||
| High | 8.1 | EPSS 0.6% | CVE-2025-66336Apache Doris MCP Server · SQL Injection High8.1EPSS 0.6%CVE Database V5CVE-2025-66336 Apache Doris MCP Server contains a SQL injection vulnerability in a metadata query path. A user-controlled database name is directly interpolated into a SQL query, and the query is executed without passing the caller's authorization context. This may allow an authenticated attacker, or an anonymous attacker if authentication is disabled, to bypass SQL security validation and access metadata outside the intended database scope. Affected users are recommended to upgrade to Doris version 0.6.1 or later, which fixes the issue. | CVE Database V5 | ||
| Medium | 5.3 | EPSS 0.4% | CVE-2026-12798BerriAI litellm · Server-Side Request Forgery Medium5.3EPSS 0.4%CVE Database V5CVE-2026-12798 A weakness has been identified in BerriAI litellm up to 1.82.2. Affected by this vulnerability is the function load_openapi_spec_async of the file litellm/proxy/_experimental/mcp_server/openapi_to_mcp_generator.py of the component MCP OpenAPI Spec Loader. This manipulation of the argument spec_path causes server-side request forgery. It is possible to initiate the attack remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure. EPSS 0.4%Vulnerability | CVE Database V5 | ||
| Medium | 5.3 | EPSS 0.4% | CVE-2026-12774BerriAI litellm · Server-Side Request Forgery Medium5.3EPSS 0.4%CVE Database V5CVE-2026-12774 A security vulnerability has been detected in BerriAI litellm up to 1.82.2. Affected by this vulnerability is the function _execute_with_mcp_client of the file litellm/proxy/_experimental/mcp_server/rest_endpoints.py of the component MCP Server Connection Testing. The manipulation leads to server-side request forgery. Remote exploitation of the attack is possible. The exploit has been disclosed publicly and may be used. The vendor was contacted early about this disclosure. EPSS 0.4%Vulnerability | CVE Database V5 | ||
| Medium | 6.5 | EPSS 0.4% | Medium6.5EPSS 0.4%CVE Database V5CVE-2026-49359 PhpWeasyPrint is a PHP library allowing PDF generation from a URL or an HTML page. Prior to version 2.6.0, `pontedilana/php-weasyprint` fetches the content of option values server-side via `file_get_contents()` when the value looks like a URL, without restricting the URL scheme. The `attachment` option of `Pdf` is the reachable sink: any value that passes `isOptionUrl()` (`filter_var(..., FILTER_VALIDATE_URL)`) is downloaded by the PHP process and embedded into the generated PDF. Because `FILTER_VALIDATE_URL` accepts `http`, `https`, `ftp`, `file` and PHP stream wrappers such as `php://`, an attacker who can influence the `attachment` value reaches both a **Server-Side Request Forgery** primitive (e.g. internal HTTP endpoints, cloud metadata) and a local file disclosure primitive (`file://`, `php://filter/...`), with the fetched bytes exfiltrated as a PDF attachment. This is the same class of issue KnpLabs/snappy patched for its `xsl-style-sheet` option in GHSA-c5fp-p67m-gq56. The library is documented as a one-to-one substitute for KnpLabs/snappy and shares the same code shape. PhpWeasyPrint version 2.6.0 contains a patch for the issue. | CVE Database V5 | ||
| –Not rated | – | AlienVault OTX General A comprehensive analysis of 339 commands issued by a French-speaking threat actor nicknamed 'Poisson' over 33 days, targeting a French automotive small business and four French individuals. The attacker utilized a multi-stage fileless attack deploying a 70-line Python keylogger to harvest banking and email credentials. The operation leveraged free-tier infrastructure including Havoc C2 framework, Backblaze B2 storage, and DuckDNS. Most significantly, the attacker installed OpenSSH and Tailscale VPN on victim machines, creating persistent access that survived C2 server takedown. When the C2 went offline for 18 days, the attacker's access remained intact through the VPN mesh, demonstrating that VPN-mesh-based persistence is actively used in real-world intrusions and that traditional C2 takedown is insufficient for remediation. | AlienVault OTX General | |||
| Medium | 4.2 | EPSS 0.2% | CVE-2026-48522SUSE kimi-cli · Security update for python-PyJWT Medium4.2EPSS 0.2%GCVE DatabaseCVE-2026-48522 PyJWKClient in the kimi-cli library allows fetching JSON Web Key Sets (JWKS) from URLs with schemes beyond HTTP and HTTPS, including file://, ftp://, and data:. This can lead to server-side request forgery (SSRF) on local files and other protocols. An attacker who can influence the jku URL and write to a local file path can forge JWT tokens that the application accepts as valid. The vulnerability affects PyJWT versions 2.11.0 and 2.12.1 and likely earlier versions. A fix is available that restricts allowed URL schemes to HTTP and HTTPS. | GCVE Database | ||
| Malicious | – | MaliciousGCVE Database The ts-big-ecro npm package is a malicious typosquatting clone of the legitimate big.js library. It impersonates the original by copying its code and metadata but injects attacker-controlled code that executes on every import or require. This code loads an undeclared and suspicious dependency, parket-helper, which can execute arbitrary code. The package's declared dependency server-parket is unrelated to its stated purpose, indicating a dependency-confusion trojan-loader pattern. Installing or running this package compromises the host system, potentially exposing all stored secrets and keys. | GCVE Database | |||
| Low | 3.5 | EPSS 0.2% | CVE-2026-12047pgAdmin 4 · Cross-site Scripting Low3.5EPSS 0.2%CVE Database V5CVE-2026-12047 HTML injection in pgAdmin 4's cloud deployment module. The verify_credentials, deploy, regions, and update-server endpoints under /rds/, /azure/, /google/, and the top-level /cloud/ blueprint propagated AWS / Azure / Google SDK exception text — and the related file-resolution and database-commit exception text — into the JSON response body (the info and errormsg fields) without HTML-encoding. The Cloud Wizard frontend rendered these strings through html-react-parser, so an attacker-influenced exception message embedded structural HTML directly into the wizard's DOM. The reported entry point is /rds/verify_credentials/. An authenticated pgAdmin user submits a crafted access_key whose value contains an <iframe/src=...> payload; AWS STS rejects the credential with an IncompleteSignature exception whose text quotes the access_key verbatim; the pgAdmin backend forwards that text into the JSON info field; the Cloud Wizard's FormFooterMessage parses it as HTML. The browser fetches the iframe's src from an attacker-controlled host, and JavaScript executing inside the cross-origin iframe writes to parent.location, redirecting the victim's pgAdmin tab. Because the injection renders inside pgAdmin's own interface, X-Frame-Options and Content-Security-Policy frame-ancestors do not mitigate it. Baseline impact is self-targeted (the same user who supplied the payload sees the injection); escalation against other authenticated users requires an additional cross-site request-forgery primitive capable of submitting the malformed credential request with a valid X-pgA-CSRFToken in the victim's browser context. The same unsanitised-error-into-JSON pattern was present across multiple sibling endpoints — Azure's check_cluster_name_availability, every Google endpoint that surfaces SDK errors (verification_ack, projects, regions, instance_types, database_versions, the verify_credentials path-resolution branches), the central /deploy endpoint that bubbles str(e) from deploy_on_rds / deploy_on_azure / deploy_on_google, and update_cloud_server which surfaces the str(e) from a failing db.session.commit — all of which are now covered. Fix HTML-escapes every external/SDK exception string at the endpoint sink via a new shared sanitize_external_text helper (HTML escape with control-character strip), promoted out of the psycopg3 driver into web/pgadmin/utils/text_sanitize.py. The Cloud Wizard frontend additionally renders its FormFooterMessage in plain-text mode for backend-derived strings, so the value is never parsed as HTML even if a future sink forgets the escape. This issue affects pgAdmin 4: from 6.6 before 9.16. | CVE Database V5 | ||
| Critical | 9.0 | EPSS 0.7% | CVE-2026-12045pgAdmin 4 · SQL Injection Critical9.0EPSS 0.7%CVE Database V5CVE-2026-12045 Read-only transaction bypass in the pgAdmin 4 AI Assistant allows an attacker who can influence database content that the assistant reads to execute arbitrary SQL with the privileges of the pgAdmin user's database role. The AI Assistant's execute_sql_query tool runs LLM-generated SQL inside a BEGIN TRANSACTION READ ONLY wrapper to prevent data modification. The LLM-supplied query was forwarded to the database driver without restriction to a single statement or to read-only verbs, so a multi-statement payload beginning with COMMIT, END, ROLLBACK, or ABORT terminated the read-only transaction and ran subsequent statements in autocommit mode. The trailing ROLLBACK then had no effect. Delivery is via prompt injection: an attacker who can write content into any object the AI Assistant may inspect (a row, a column value, a comment) can cause the LLM to emit the multi-statement payload as a tool call. With ordinary write privileges on the pgAdmin user's role the attacker can perform unauthorised data modification. When the pgAdmin user's role is a PostgreSQL superuser or holds pg_execute_server_program, the chain extends to remote code execution on the database server host via COPY ... TO PROGRAM. Fix validates the LLM-supplied query up front: it must parse to exactly one non-empty / non-comment statement whose leading real token (after stripping whitespace, comments, and punctuation) is one of SELECT, WITH, EXPLAIN, SHOW, VALUES, or TABLE. Transaction-control verbs, DML, DDL, CALL, COPY, DO, SET/RESET, and everything else are rejected before any database work happens. PostgreSQL's READ ONLY mode continues to backstop data-modifying CTEs, EXPLAIN ANALYZE on writes, and volatile side effects. This issue affects pgAdmin 4: from 9.13 before 9.16. | CVE Database V5 | ||
| Critical | 10.0 | EPSS 0.6% | CVE-2026-49257startreedata mcp-pinot · Missing Authentication Critical10.0EPSS 0.6%CVE Database V5CVE-2026-49257 mcp-pinot is a Python-based Model Context Protocol (MCP) server for interacting with Apache Pinot. In versions 3.0.1 and below, mcp-pinot defaults to running an HTTP MCP server bound to 0.0.0.0:8080 with no authentication enabled. All MCP tools, including SQL query execution, schema creation, and table-config mutation, are reachable by any network-adjacent caller. The server proxies these calls using server-side Pinot credentials, producing a confused-deputy condition that yields full read/write access to the configured Pinot cluster. This issue has been fixed in version 3.1.0 | CVE Database V5 | ||
| High | 7.5 | EPSS 0.4% | CVE-2026-48818Kludex starlette · Server-Side Request Forgery (SSRF) High7.5EPSS 0.4%CVE Database V5CVE-2026-48818 ### Summary When serving static files on Windows, `StaticFiles` resolves the requested path with [`os.path.realpath`](https://docs.python.org/3/library/os.path.html#os.path.realpath). If a UNC path (such as `\\attacker.com\share`) reaches the resolver, `realpath` causes the process to open a connection to the remote host over SMB (port 445). This is a server-side request forgery (SSRF) that leaks the service account's NTLMv2 credentials to the attacker-controlled host, which can then be cracked offline or relayed to other hosts. ### Details `StaticFiles.lookup_path()` joins the requested path onto the served directory and calls [`os.path.realpath`](https://docs.python.org/3/library/os.path.html#os.path.realpath) on the result before checking containment with [`os.path.commonpath`](https://docs.python.org/3/library/os.path.html#os.path.commonpath). On Windows, a UNC path is absolute, so [`os.path.join`](https://docs.python.org/3/library/os.path.html#os.path.join) discards the served directory and `realpath` resolves the bare UNC path, triggering the outbound SMB connection and NTLM authentication before the containment check rejects the path. The HTTP response is a benign 404, but the credential disclosure has already happened. POSIX systems are not affected. This only affects the default configuration (`follow_symlink=False`), which uses [`os.path.realpath`](https://docs.python.org/3/library/os.path.html#os.path.realpath). The `follow_symlink=True` branch uses [`os.path.abspath`](https://docs.python.org/3/library/os.path.html#os.path.abspath), which performs no I/O. ### Impact Applications running on Windows that serve files with `StaticFiles` (directly, or via a framework built on Starlette such as FastAPI) in the default configuration are affected. `StaticFiles` is typically unauthenticated, so any client can trigger the SMB connection and leak the service account's NTLMv2 hash. A secondary impact is discovering internal hosts reachable over SMB by timing responses for valid versus invalid addresses. ### Mitigation Applications not running on Windows are not affected. On Windows, serving static files through a dedicated web server (such as nginx or IIS) instead of `StaticFiles` avoids the issue. Blocking outbound SMB (port 445) from the application host prevents the credential disclosure even if a UNC path is resolved. | CVE Database V5 | ||
| High | 7.7 | EPSS 0.4% | CVE-2026-54017Open WebUI · Path traversal / SSRF in terminal server proxy via encoded path traversal High7.7EPSS 0.4%GCVE DatabaseCVE-2026-54017 An authenticated user with access to a terminal server in open-webui can exploit a path traversal and SSRF vulnerability via crafted encoded path segments. The vulnerability arises because the terminal-server reverse proxy does not fully sanitize user-controlled path inputs before forwarding them, allowing traversal outside intended scopes. A double-encoding technique bypasses the existing sanitization, enabling access to unintended files or internal services. This affects open-webui versions prior to 0.9.6. | GCVE Database | ||
| Critical | 9.3 | EPSS 0.9% | CVE-2026-53805nv-tlabs GEN3C · Deserialization of Untrusted Data Critical9.3EPSS 0.9%CVE Database V5CVE-2026-53805 NVIDIA Spatial Intelligence Lab's (SIL) GEN3C contains an unauthenticated remote code execution vulnerability in the inference API server where the /request-inference and /seed-model endpoints deserialize raw HTTP request bodies using Python's pickle.loads() without authentication or input validation. Attackers can supply a crafted payload containing a __reduce__ gadget to the inference API port to achieve remote code execution as the inference process. EPSS 0.9%Vulnerability | CVE Database V5 | ||
| Malicious | – | MaliciousGCVE Database --- _-= Per source details. Do not edit below this line.=-_ ## Source: amazon-inspector (c7d67bcaf42f4ef050a6b824dfbc2f711d6820f7b3b1140d211110f390e512fa) The package ships several files whose names and contents suggest a non-trivial install-time and runtime footprint: a postinstall script (scripts/postinstall-agent.mjs) that performs network I/O including a ping/GET against an identifier-bearing endpoint, a relay server (dist/relayServer.js) with multiple ping invocations, a Discord relay upload module (dist/discordRelayUpload.js) combining base64 decoding with POST calls, a Hugging Face credentials module (dist/hfCredentials.js) using base64-decoded values, and a secret-scan startup audit (dist/secretScan/agentStartupAudit.js) that fetches huggingface.co endpoints. The combination of a postinstall agent, a Discord upload path, base64-encoded data in deployment defaults, and credential/secret handling is consistent with either (a) a legitimate ML/agent tooling package that talks to Hugging Face and uses Discord for notifications, or (b) a credential-relay/exfiltration tool. The evidence available here is keyword-level pattern matches without traced execution paths, so intent cannot be determined from this data alone. A human reviewer should inspect scripts/postinstall-agent.mjs to confirm what runs on `npm install`, dist/discordRelayUpload.js to confirm whether installer-side data is shipped to Discord, and dist/hfCredentials.js to confirm whose credentials are read. ## Source: ghsa-malware (95b8b8f54c9c2a4fa965e4a7fe7ea9c8f8ef1b5bcdd4cc1da7354d0ce2e8842f) Any computer that has this package installed or running should be considered fully compromised. All secrets and keys stored on that computer should be rotated immediately from a different computer. The package should be removed, but as full control of the computer may have been given to an outside entity, there is no guarantee that removing the package will remove all malicious software resulting from installing it. | GCVE Database | |||
| High | 8.4 | EPSS 0.4% | CVE-2025-26240JazzCore python-pdfkit · Classic Buffer Overflow High8.4EPSS 0.4%CVE Database V5CVE-2025-26240 In JazzCore python-pdfkit 1.0.0, the from_string method enables the execution of JavaScript code within the context of the server application and the exfiltration of local files. | CVE Database V5 | ||
| Medium | 6.8 | EPSS 0.4% | CVE-2026-48782pydantic-ai · Server-Side Request Forgery (SSRF) Medium6.8EPSS 0.4%CVE Database V5CVE-2026-48782 ## Summary When an application using Pydantic AI opts a URL into `force_download='allow-local'` (which disables the default block on private/internal IPs) **and runs on a network that routes the affected IPv6 transition forms (NAT64- or ISATAP-configured networks)**, the cloud-metadata blocklist could be bypassed by encoding the metadata IP in an IPv6 transition form that the previous fix did not decode — IPv4-compatible IPv6 (`::a.b.c.d`), the NAT64 RFC 8215 local-use prefix (`64:ff9b:1::/48`), operator-chosen NAT64 prefixes, or ISATAP. The IPv6 wrapper is then delivered to the underlying IPv4 metadata endpoint, exposing cloud IAM short-term credentials. **The bypass is exploitable only in environments whose network actually routes these forms** — NAT64-configured networks (IPv6-only or dual-stack-with-NAT64 deployments, including some Kubernetes setups) for the NAT64 variants, or networks with an ISATAP tunnel for ISATAP. A standard dual-stack cloud VM or container does not route them and is not affected in practice. The IPv4-compatible and Teredo variants are deprecated and addressed as defense-in-depth. This is an incomplete fix of [GHSA-cqp8-fcvh-x7r3](https://github.com/pydantic/pydantic-ai/security/advisories/GHSA-cqp8-fcvh-x7r3) / [CVE-2026-46678](https://nvd.nist.gov/vuln/detail/CVE-2026-46678) (itself a follow-up to [CVE-2026-25580](https://nvd.nist.gov/vuln/detail/CVE-2026-25580)). The prior remediation decoded only IPv4-mapped IPv6, 6to4, and the NAT64 well-known prefix; the metadata guarantee did not hold for the remaining transition forms. ## Severity **MEDIUM** — `CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N` = **6.8** Same impact metrics and narrow attack surface as the parent advisory (AC:H): exploitation requires the application to have opted into `allow-local` on a URL influenced by untrusted input, and the NAT64/ISATAP variants additionally require the deployment network to route those forms. **CWE-918**: Server-Side Request Forgery (SSRF) ## Affected Versions | Package | Vulnerable | Patched | |---|---|---| | `pydantic-ai` | `>= 1.56.0, < 1.102.0`; `>= 2.0.0b1, < 2.0.0b3` | `1.102.0`; `2.0.0b3` | | `pydantic-ai-slim` | `>= 1.56.0, < 1.102.0`; `>= 2.0.0b1, < 2.0.0b3` | `1.102.0`; `2.0.0b3` | These transition forms have not been decoded since SSRF protection was introduced in `1.56.0`. ## Who Is Affected Users are affected **only if** their application explicitly opts a `FileUrl` (`ImageUrl`, `AudioUrl`, `VideoUrl`, `DocumentUrl`) into `force_download='allow-local'` on a URL that is, or could be, influenced by untrusted input. Beyond that precondition, the affected encodings only reach a metadata endpoint in environments whose network actually routes them. The broadly-routable IPv4-mapped form was addressed in `1.99.0` (CVE-2026-46678); the additional forms addressed here require a **NAT64-configured network** (IPv6-only or dual-stack-with-NAT64 deployments, including some Kubernetes setups) for the NAT64 variants, or an **ISATAP tunnel** for the ISATAP variant. The IPv4-compatible and Teredo forms are deprecated and not routed by modern stacks; they are addressed as defense-in-depth. Most deployments on a standard dual-stack cloud VM or container are therefore not exploitable in practice, but the fix restores the "always blocked" guarantee for the environments that are. Users are **not** affected if they use any of the bundled integrations to ingest user input, because they do not propagate `force_download` from external data: - `Agent.to_web` / `clai web` - `VercelAIAdapter` - `AGUIAdapter` / `Agent.to_ag_ui` Applications that only download from developer-controlled URLs are not affected. ## Remediation Upgrade to `1.102.0` or later (or `2.0.0b3` or later on the 2.0 pre-release line). The cloud-metadata and private-IP blocklists now decode the embedded IPv4 of every standardized IPv6 transition form before evaluating it — IPv4-mapped, IPv4-compatible, 6to4, NAT64 across all prefix lengths (including the RFC 8215 local-use prefix and operator-chosen prefixes), ISATAP, and Teredo. The set of always-blocked cloud metadata/credential endpoints has also been expanded across providers. ## Workaround for Unpatched Versions Avoid passing `force_download='allow-local'` on any URL that could be influenced by untrusted input. If developers must, resolve the hostname themselves and validate the result against their own metadata blocklist — including IPv6 transition forms — before constructing the `FileUrl`. ## Credits Reported by [@SnailSploit](https://snailsploit.com). | CVE Database V5 |
Page 50 of 79