## Summary Hysteria's UDP relay treats the destination address as packet-scoped, but ACL and outbound policy are applied only once when a new UDP session is created. After an authenticated client opens a UDP session using an allowed first destination, later packets in the same `Session ID` can be sent to different destinations without re-running ACL evaluation. This allows an authenticated user to bypass server-side UDP ACL rules and reach localhost or RFC1918/private-network UDP services from the server's network perspective, even when those destinations are explicitly rejected by ACL. Verified on current HEAD at commit `64c396385631579598cc29d5561bff98c439772f`. ## Why this is a security issue This report is not based on the assumption that one UDP session must be bound to one destination. The protocol and official client both support per-packet destinations: - `PROTOCOL.md:93-107` defines each `UDPMessage` as carrying its own `Addr` field. - `core/client/udp.go:52-62` exposes `Send(data, addr)`, allowing the same UDP session to send to arbitrary addresses. The problem is that the security-relevant destination is packet-scoped, while ACL and outbound authorization are cached at session scope. This is also not a `RequestHook`-bypass claim. I understand `RequestHook` is first-packet-oriented. The broader issue is that operator-configured ACL policy intended to block UDP destinations is not enforced on later packets within the same session. Because the ACL documentation is presented as the mechanism for handling or blocking client requests, and includes examples of denying `udp/443` and private network CIDRs, operators can reasonably rely on ACL as a UDP egress security boundary. This boundary can currently be bypassed by reusing a previously authorized UDP session. ## Root cause The relevant flow appears to be: - `core/server/udp.go:280-299`: when a new session is created, the first destination is passed through `m.io.Hook(...)`, logged, and then `m.io.UDP(addr)` is called once to create the outbound UDP connection. - `core/server/server.go:397-398`: `m.io.UDP(addr)` delegates to `io.Outbound.UDP(reqAddr)`. - `app/cmd/server.go:1187-1190`: resolver, ACL, and actual outbounds are intentionally chained through the `Outbound` interface. - `core/server/udp.go:125`: the initial outbound connection is created only from the first packet via `DialFunc(firstMsg.Addr, firstMsg.Data)`. - `core/server/udp.go:92-111`: later packets in the same session take the current packet address and directly call `e.conn.WriteTo(dfMsg.Data, addr)` without re-running ACL or outbound policy evaluation. In other words, destination selection is packet-scoped, but authorization is session-scoped. ## Impact Any authenticated client that is allowed to use UDP relay can: - open one UDP session using an allowed first destination; - reuse the same session to send packets to destinations that ACL should reject; - reach UDP services on `127.0.0.1` or on RFC1918/private-network addresses from the server's network perspective. In real deployments, this can expose internal-only UDP services such as: - internal DNS resolvers; - service discovery endpoints; - telemetry or metrics listeners; - local administrative daemons; - application-specific UDP services intended to be reachable only from localhost or the internal network. This breaks the server's documented ACL-based UDP egress restrictions. ## Reproduction Two cases were reproduced with integration tests. ### Case 1: localhost bypass ACL: ```text direct(127.0.0.1, udp/<allowedPort>) reject(127.0.0.1/32) ``` Steps: 1. Start one UDP echo service on `127.0.0.1:<allowedPort>`. 2. Start another UDP echo service on `127.0.0.1:<blockedPort>`. 3. Connect an authenticated Hysteria client and create one UDP session. 4. Send a packet to the allowed loopback destination to establish the session. 5. Reuse the same UDP session and send a packet to the blocked loopback destination. Observed result: - The second packet is relayed successfully and the blocked loopback service replies. Expected result: - The second packet should be rejected because `127.0.0.1/32` is denied by ACL. ### Case 2: private-network bypass ACL: ```text direct(127.0.0.1, udp/<allowedPort>) reject(10.0.0.0/8) ``` or the corresponding local RFC1918 range, such as `192.168.0.0/16` or `172.16.0.0/12`. Steps: 1. Start one UDP echo service on `127.0.0.1:<allowedPort>`. 2. Start another UDP echo service on a real RFC1918 address of the server host. 3. Connect an authenticated Hysteria client and create one UDP session. 4. Send a packet to the allowed loopback destination first. 5. Reuse the same UDP session and send a packet to the RFC1918 destination. Observed result: - The private-address packet is relayed successfully and receives a reply. Expected result: - The packet should be rejected by ACL. ## PoC and local evidence A local integration test file was added during verification: - `core/internal/integration_tests/udp_private_acl_bypass_test.go` The two tests are: - `TestClientServerUDPACLBYPASSLoopback` - `TestClientServerUDPACLBYPASSPrivateIPv4` They can be executed with: ```bash go test ./core/internal/integration_tests -run 'TestClientServerUDPACLBYPASS(Loopback|PrivateIPv4)' -count=1 ``` The tests pass locally and demonstrate that a destination blocked by ACL becomes reachable after the session is established with an allowed first destination. ## Suggested fixes Any of the following would address the issue: 1. Re-evaluate ACL and outbound policy for every defragmented UDP packet before forwarding it with `WriteTo`. 2. Alternatively, enforce a single immutable destination per UDP session and reject destination changes after the first packet. 3. Ensure logging and policy hooks are aligned with the chosen model so that policy enforcement and observability reflect the real per-packet destination. ## Severity assessment Suggested CVSS v3.1 vector: `CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:L/I:L/A:L` This reflects a network-reachable issue with low attack complexity, requiring only an authenticated client, no victim interaction, and allowing impact beyond the proxy process by exposing localhost and internal-network UDP resources from the server's trust boundary.
PoC: mikrotrick-poc
CVE-2026-67276 RouterOS SSH public-key authentication bypass lab PoC
PoC: xiaomi15-dada-cve-2026-64560
Device-bound CVE-2026-64560 adaptation for Xiaomi 15 dada OS4.0.0.8
PoC: cve-2026-32475-elementor-pro-lab
A/B Docker lab + PoC for CVE-2026-32475 (Elementor Pro Forms unauthenticated arbitrary file upload -> RCE via validation/move loop desync)
PoC: KeySniper
**CVE-2026-18963** — unauthenticated Keycloak account takeover via the reset-credentials flow.
PoC: CVE-2026-58138
CVE-2026-58138
PoC: CVE-2026-41940
cPanel & WHM - Authentication Bypass via Session-File CRLF Injection
PoC: CVE-2024-12356
Unauthenticated RCE detector + RCA for BeyondTrust Remote Support / PRA (CVE-2024-12356 + CVE-2025-1094)
PoC: CVE-2026-85046
CVE-2026-85046
PoC: gha-lab-360f77d0d4
Authorized security-research lab: reproduction of CVE-2026-39866 (GHSA-9prc-pp2c-3427) — workflow_dispatch input template injection in .github/workflows/release_update.yml of LawnchairLauncher/lawnchair @ b089bae8c007f36a8ce0346725182a107d97cd05. Snapshot pinned to the vulnerable commit; owner-gated sign-info step retargeted for the lab.
PoC: log4shell-exploitation-lab
CVE-2021-44228 Log4Shell reproduced end to end: exploitation through remediation
PoC: GitLab-CVE-2023-7028
A mock app for the GitLab CVE-2023-7028, which allow multile email adresses when ordering a password reset.
PoC: CVE-2021-3030
Advisory: Cute Editor 6.4 reflected XSS via 'Theme' parameter in colorpicker_more.aspx
PoC: CVE-2026-27876
Grafana SQL Expressions Arbitrary File Write to RCE
PoC: CVE-2026-28956-jxl-messages-surface
JPEG XL auto-decodes in the iOS Messages preview path — delivery-surface finding for CVE-2026-28956 (AppleJPEGXL), with patch-diff attribution (libjxl 0.10.4->0.10.5) and an honest reliability check on the public PoC.
PoC: CVE-2026-73570
Zimbra Collaboration Suite RCE — SMTP log poisoning → swatchdog → OS Command Injection (CVSS 8.9, CISA KEV)
PoC: CVE-2020-10770-keycloak-exploit-poc
Keycloak Blind SSRF POC
PoC: CVE-2026-1529-Keycloak-Exploit-Tool
Keycloak: Unauthorized organization registration via improper invitation token validation
PoC: CVE-2026-18963-keycloak
CVE-2026-18963 — Keycloak reset-credentials bypass -> Account Takeover
PoC: CVE-2026-64747
Root cause + macOS reachability PoC for CVE-2026-64747 (AppleAVE2 kext buffer overflow, fixed 26.6 / 905.40.1). Fully reversed AppleAVE2UserClient wire protocol, mode-5 LRB overflow math, IOKit PoC driving the configure path.
PoC: CVE-2026-64705
Root cause + PoC for CVE-2026-64705 (macOS HFS xattr kernel heap overflow, fixed 14.8.7). Weaponized HFS+ image: unbounded bcopy loop -> kernel heap overflow -> panic on pre-fix systems; validator rejection on patched. Kext diff, mechanism, rebuild recipe.
PoC: CVE-2026-78938
Root cause analysis + working R/W exploit for CVE-2026-78938 (V8 TurboFan CheckMaps instance-migration type confusion, Chrome 152, exploited in the wild). Crash PoC + addrof/fakeobj/arbitrary R/W over the compressed heap.
PoC: Jozini-network-scanner
# Jozini Network Scanner Built in Termux at KwaQondile Library, Jozini KZN Tools: - scanner.py: Port scanner with banner grabbing (20 ports + report saving) - cve_check.py: Maps RouterOS version to known CVEs Finding: MikroTik RouterOS 6.46.8 vulnerable to CVE-2020-2021 (Critical) Author: [Your Name] - Aspiring Pentester
PoC: CVE-2026-52774-YESWIKI-XSS
a reflected XSS vulnerability in YesWiki's Bazar widget handler.
PoC: netty-http2-check
CVE-2025-55163 / CVE-2026-56819: offline checker for the 7 netty-codec-http2 CVEs. Tells you which ones you are exposed to, and the one version that fixes all seven (4.1.136.Final / 4.2.16.Final) - written on none of the advisories. Does not scan pom.xml on purpose: WebFlux pulls it in transitively.
PoC: CVE-2026-0920
A PoC exploit for CVE-2026-0920 - LA-Studio Element Kit / Unauthenticated Privilege Escalation
PoC: CVE-2026-84645
Jenkins PersistenceRoot Deserialization RCE (SECURITY-3972) — PoC & analysis. Requires Item/Configure; affects weekly <= 2.579 / LTS <= 2.568.2
PoC: cyberthreat_DBSproject
threat = { "id": "CVE-2026-0001", "title": "Apache HTTP Server Remote Code Execution", "vendor": "Apache", "product": "HTTP Server", "description": "A vulnerability in Apache HTTP Server allows remote attackers to execute arbitrary code.", "cvss": 9.8, "kev": True, "published": "2026-06-30" }
PoC: CVE-2026-6471
CVE-2026-6471
PoC: CVE-2026-75865
Unauthenticated arbitrary file upload -> RCE in WPLP Cookie Consent (gdpr-cookie-consent) <= 4.4.1 - technical write-up and PoC
PoC: CVE-2026-32475
CVE-2026-32475 PoC : Elementor Pro Unauthenticated Arbitrary File Upload to RCE
PoC: CVE-2023-42793-TeamCity-Unauthenticated-RCE
A PoC and automated version detection/exploit tool for JetBrains TeamCity Authentication Bypass & RCE (CVE-2023-42793).
PoC: cve-2026-6471-postgres-logical-decoding-dlopen
postgres CVE-2026-6471 Exploit
PoC: gpgsm-cve-2026-57062-cms-gcm-short-tag
gpgsm CVE-2026-57062 exploit POC
PoC: CVE-2025-4255---Buffer-Overflow
Exploit Framework for CVE-2025-4255
PoC: gha-lab-4a8fad8536
Security-research lab reproducing CVE-2026-39382 (GHSA-5jxf-vmqr-5g82): command injection in dbt-labs reusable workflow open-issue-in-repo.yml, driven by a dbt-core-style docs-issue.yml caller
PoC: gha-lab-ed7a1740c4
Security-research lab: controlled reproduction of GHSA-3g6g-gq4r-xjm9 / CVE-2026-35580 (GitHub Actions workflow_dispatch input shell injection) against a pinned snapshot of NationalSecurityAgency/emissary
PoC: gha-lab-85f022290a
Research lab reproduction of CVE-2026-34243 (GHSA-r4fj-r33x-8v88): command injection via issue_comment.body in .github/workflows/comment.yaml — snapshot of njzjz/wenxian@ca4e04de86aa970c0e3cb1c7f2bd103d339fbe51
PoC: gha-lab-9b5e3ccfbe
Security-research lab: reproduction of CVE-2026-33475 (GitHub Actions script injection via PR branch name in deploy-docs-draft.yml), snapshot of langflow-ai/langflow
PoC: research-cve-2026-85649
[MIRROR] The CVE-2026-85649 Security Research Publication.
PoC: gha-lab-61c59f4acb
Security-research lab: controlled reproduction of CVE-2026-33075 (pwn request in labring/FastGPT preview-image workflow, pull_request_target + checkout-of-fork + privileged buildx push)
PoC: gha-lab-3f1ff30e9c
Authorized security-research lab reproducing CVE-2026-31852 (jellyfin/jellyfin-ios pull_request_target pwn in code-quality.yml) — isolated snapshot, not the upstream project
PoC: gha-lab-ca4fa82ac5
Security-research lab: reproduction of CVE-2026-29075 (GHSA-3j55-5q6x-2h48) in mesa/mesa benchmarks.yml pull_request_target workflow — single-commit snapshot for authorized vulnerability reproduction.
PoC: gha-lab-6c3094af9e
Authorized security-research lab reproducing CVE-2026-27941 (pwn request in pull_request_target workflows) — snapshot of openlit/openlit
PoC: gha-lab-a7f6217d26
Security-research reproduction of CVE-2026-27938 / GHSA-4q9f-mjxf-rx7x (GitHub Actions expression injection in release workflows) — snapshot of wp-graphql/wp-graphql at b216fe22f3a119f256511ec7353f536fee6886ac
PoC: cve-2026-19900-PoC
cve-2026-19900-PoC
PoC: CVE-2026-85769
Heap out-of-bounds read in libtpms TPM 2.0 state deserialization — CVE-2026-85769
PoC: CVE-2026-19632
Unauthenticated account takeover PoC for TranslatePress Multilingual <= 3.3.1 (WordPress)
PoC: CVE-2026-11613
Divi Ajax Filter <= 5.1.2 Unauthenticated Local File Inclusion via 'custom_loop_template'
PoC: gha-lab-25b7988758
Authorized security-research reproduction of CVE-2026-27701 / GHSA-xh9w-5859-x97j (live-codes/livecodes @ 8017e01): untrusted PR title interpolated into i18n-update-pull github-script block.
PoC: copy-fail-CVE-2026-31431-cpp
https://github.com/theori-io/copy-fail-CVE-2026-31431 but ported to c++ for fun
PoC: CVE-2026-83548-checker
Non-intrusive detector for SonicWall SMA 1000 exposure to CVE-2026-83548/-83549 (version/patch-state check; no exploitation)
PoC: gha-lab-b16a4f3554
Security-research lab: CVE-2026-24480 pull_request_target pre-commit RCE in qgis/QGIS (snapshot at vulnerable commit)
PoC: Yordam-Kutuphane-Otomasyonunda-Coklu-HTML-Enjeksiyonu
CVE-2026-77818 - Yordam Kütüphane Otomasyon Sistemi - Üç ayrı noktada yansıtılmış HTML enjeksiyonu, form action ele geçirme ve kimlik bilgisi hırsızlığı (CWE-79)
PoC: jsherp-user-info-idor
VulDB advisory: jshERP authenticated /user/info IDOR and password-digest replay after CVE-2025-60800
PoC: gha-lab-7927d7d06f
Security-research lab reproducing CVE-2026-22869 (pwn) — arbitrary code execution in privileged pull_request_target run via npx local-bin hijack, snapshot of eigent-ai/eigent @ 2a406536
PoC: cve-2026-31431
PoC for CVE-2026-31431
PoC: gha-lab-b5c1313658
Authorized security-research lab reproducing CVE-2026-1699 (pwn request in preview.yml) — snapshot of eclipse-theia/theia-website
PoC: CVE-2026-63077
CVE-2026-63077 - Unauthenticated RCE exploit for JetBrains TeamCity via Agent Polling Deserialization. Supports mass scanning, multi-threading, and interactive shell. For authorized security testing only.
PoC: CVE-2026-6471
CVE-2026-6471 - Draft or TODO
PoC: CVE-2026-73554
CVE-2026-73554 - Draft or TODO
PoC: CVE-2026-19516
CVE-2026-19516
PoC: gha-lab-51c6b6d0a0
Lab reproducing CVE-2025-67727 (parse-community/parse-server ci-performance.yml pull_request_target RCE at e78e58d) — authorized security research
PoC: gha-lab-6904b2ccbe
Security-research lab: reproduction of CVE-2025-61584 (GHSA-9g7x-737f-5xpc) — command injection via github.head_ref in pull_request_target workflow (.github/workflows/pr.yml)
PoC: CVE-2026-85046-Patch-confusion-zero-day-vulnerability-in-Google-Chrome-s-V8-engine
Conceptual C++ patch and structural analysis for CVE-2026-85046, a critical type confusion zero-day vulnerability in Google Chrome's V8 engine
PoC: cve-disclosures
CVE-2024-57551, CVE-2024-57552, CVE-2024-57553 advisories by Aman Bahiniya
PoC: unit-01-severity-vs-risk-reflection
cve-2026-25524 Holds no customer payment data, no monitoring in place, monitored 24/7 The CVSS score is technically serious, but it doesn't tell how exposed it is, weather our existing defenses would stop or contain an attack. We should confirm the vulnerable component is reachable by untrust input in our environment.
PoC: gha-lab-d14c91f1bb
Security-research lab: reproduction of CVE-2025-58371 (GitHub Actions command injection via PR title in Discord PR Notifier), snapshot of RooCodeInc/Roo-Code @ 08a825f9bb0086a88cff5a79b9af4731bba7d076
PoC: thymeleaf-check
Offline checker for Thymeleaf CVE-2026-40477 / CVE-2026-41901 — tells you which of the two CVSS 9.0 SSTI flaws you are exposed to, and whether your version line has a fix at all (3.0.x: it does not)
PoC: CVE-2024-36058
CVE-2024-36058 — Authenticated Time-Based Blind SQL Injection in Koha Library Software < 22.05.22 (opac-sendbasket.pl). Advisory + PoC by Hacklantic.
PoC: CVE-2024-36057
CVE-2024-36057 — Authenticated OS Command Injection in Koha Library Software < 22.05.22 (upload-cover-image.pl). Advisory + PoC by Hacklantic.
PoC: gha-lab-aa1cbc9bcf
Authorized security-research reproduction of CVE-2025-54594 (GHSA-588g-38p4-gr6x): privileged issue_comment-triggered canary release workflow checking out untrusted fork code and running its npm scripts with GITHUB_TOKEN/NPM_TOKEN in env. Snapshot of callstackincubator/react-native-bottom-tabs @ d765b1f695762490327dcb8f6a2f17542cf0abdb.
PoC: CVE-2026-82329-poc
CVE-2026-82329 Poc
PoC: CVE-2025-34158-CVE-2020-5741
CVE-2025-34158, CVE-2020-5741 - Draft or TODO
PoC: gha-lab-ba981941f0
Security-research lab reproducing CVE-2025-54430 (GHSA-wrg3-xqw8-m85p): secrets exfiltration via issue_comment-triggered Benchmark Bot in dedupeio/dedupe. Snapshot of dedupeio/dedupe@54ecfe77d41390da66899596834a2bde3712c966.
PoC: gha-lab-f894926966
Authorized security-research reproduction lab for CVE-2025-54415 (GHSA-g5hx-xv45-9whg): astronomer/dag-factory snapshot at 464c75a — pull_request_target head-SHA checkout executes attacker-controlled hatch scripts in base-repo context
PoC: gha-lab-6926364d94
Security research lab reproducing CVE-2025-53546 (GHSA-h87r-5w74-qfm4): pull_request_target arbitrary code execution in RSSNext/Folo's auto-fix lint workflow — authorized, isolated reproduction
PoC: CVE-2025-8518
CVE-2025-8518 - Draft or TODO
PoC: gha-lab-3b0a828a69
Security-research lab reproducing CVE-2025-53104 (GHSA-432r-9455-7f9x): command injection in discussion-to-slack.yml of gluestack/gluestack-ui
PoC: gha-lab-e8902eccd3
Security research lab: reproduction of CVE-2025-52467 (pgai pull_request_target workflow code execution / GITHUB_TOKEN exfiltration) — snapshot of timescale/pgai
PoC: tomcatfileread
CVE-2020-1938 (Ghostcat) Tomcat AJP file read/file include PoC with python3 port
PoC: CVE-Chamilo-LMS
CVE-2026-61578, CVE-2026-61582, CVE-2026-61583, CVE-2026-61584, CVE-2026-61585, CVE-2026-61587, CVE-2026-61600, CVE-2026-61601, CVE-2026-61602, CVE-2026-70647, CVE-2026-70648 - Draft or TODO
PoC: gha-lab-2f775f277c
Authorized lab reproduction of CVE-2025-47928 (spotipy-dev/spotipy pull_request_target secrets exfiltration) — snapshot at vulnerable commit 4f5759d
PoC: CVE-2026-31787
Linux kernel double free in Xen privcmd driver
PoC: gha-lab-fb6df3d456
Authorized security-research lab reproducing CVE-2025-46820 (GHSA-cwj7-6v67-2cm4): GITHUB_TOKEN persisted into publicly downloadable CI artifacts in phpgt/Dom. Snapshot of phpgt/Dom @ b73d7e8.
PoC: CVE-2026-20212
CVE-2026-20212 - Draft or TODO
PoC: CVE-2026-56718
AJCloud AJY IPC Firmware Path Traversal via jdbhttpd
PoC: psa-2026-00043-recovery
Recovery notes for proxmox advisory ID: PSA-2026-00043-1 (CVE-2023-54391)
PoC: gha-lab-ba8e0c4217
Authorized security-research lab: reproduction of CVE-2024-42370 / GHSA-4hq2-rpgc-r8r7 (env injection in docs-preview.yml) — snapshot of litestar-org/litestar@18d84d84
PoC: CVE-2026-65643-PoC-Toolkit
🧰 CVE-2026-65643 – cPanel Domain Parking RCE Toolkit (CVSS 8.7) | Red/Blue Team suite for unpatched cPanel & WHM 11.x (110,134,136,138). 2 tools: Full Exploit (reverse shell, webshell, persistence, root passwd, file R/W, mass scan, Tor), Blue Team PoC (detection, reporting, audit). w/Python. 🦾 Only Use Ethically, Stay Legal <3
PoC: CVE-2026-4813
PoC for CVE-2026-4813
PoC: cve-2026-75604
Research lab and exploit chain for CVE-2026-75604: path traversal in the Next.js incremental cache, to RCE on Windows.
PoC: CVE-2026-82329
CVE‑2026‑82329 is a critical authentication bypass in JFrog Artifactory (CVSS 9.8) allowing unauthenticated attackers to obtain full administrative privileges. Actively exploited in the wild. Affects self‑hosted versions before patches. PoC for authorized testing only.
PoC: CVE-2026-52810
CVE-2026-52810 - Draft or TODO
PoC: iOS26.6-CVE-2026-64788
CVE-2026-64788 PoC — IOGPUFamily Use-After-Free (iOS 26.6 / 23G71)
PoC: CVE-2026-80428
CVE-2026-80428 PoC
PoC: iOS26.6-CVE-2026-65343
CVE-2026-65343 PoC — AppleKeyStore OOB read → KASLR defeat (iOS 26.6 / 23G71)
PoC: CVE-2026-80428
CVE-2026-80428 PoC
PoC: gha-lab-b1fe4918c0
Authorized security-research lab: reproduction of CVE-2025-32958 (GHSA-8c7v-vccv-cx4q) — GITHUB_TOKEN leaked into workflow artifacts by Adept's remoteBuild.yml (snapshot of AdeptLanguage/Adept @ 6a64554)
PoC: CVE-2026-83548-SonicWall-SMA1000-Analysis
Vulnerability Analysis of CVE-2026-83548 affecting SonicWall SMA1000 security systems.
PoC: CVE-2024-21546
This repository contains security assessment tooling, detection templates, and an automated exploit toolkit for identifying and exploiting Unauthenticated Remote Code Execution (RCE) in applications utilizing the `UniSharp/laravel-filemanager` package (Versions `< 2.9.1`).
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:L/I:L/A:L
Get alerted for CVEs like this
Register your stack and get notified within minutes when a matching CVE drops.
Start monitoring free