{"metadata": {"vulnerability": "CVE-2026-33359", "count": 15, "kinds": {"milestone": 4, "sighting": 2, "kev": 0, "other": 9}, "linked_omitted": 0}, "data": [{"timestamp": "2026-03-19T00:27:05.987Z", "date_only": false, "value": "2026-03-19T00:27:05.987Z", "kind": "milestone", "label": "Reserved", "source": "cvelistv5", "record": "CVE-2026-33359", "field": "cveMetadata.dateReserved"}, {"timestamp": "2026-05-11T16:00:00Z", "date_only": false, "value": "2026-05-11T16:00:00.000Z", "kind": "other", "label": "Made public", "source": "cvelistv5 / runZero", "record": "CVE-2026-33359", "field": "containers.cna.datePublic", "metric_role": "disclosure", "metric_basis": "recorded_event"}, {"timestamp": "2026-05-11T16:03:21.535Z", "date_only": false, "value": "2026-05-11T16:03:21.535Z", "kind": "milestone", "label": "Published", "source": "cvelistv5", "record": "CVE-2026-33359", "field": "cveMetadata.datePublished", "metric_role": "disclosure", "metric_basis": "recorded_event"}, {"timestamp": "2026-05-11T16:03:21.535Z", "date_only": false, "value": "2026-05-11T16:03:21.535Z", "kind": "other", "label": "Updated", "source": "cvelistv5 / runZero", "record": "CVE-2026-33359", "field": "containers.cna.providerMetadata.dateUpdated"}, {"timestamp": "2026-05-11T17:16:30Z", "date_only": false, "value": "2026-05-11T17:16:30Z", "kind": "other", "label": "Nvd published at", "source": "github \u00b7 GHSA-FXWQ-QQHR-34Q4", "record": "GHSA-FXWQ-QQHR-34Q4", "field": "database_specific.nvd_published_at", "linked_record": true}, {"timestamp": "2026-05-11T17:16:30.843Z", "date_only": false, "value": "2026-05-11T17:16:30.843", "kind": "milestone", "label": "Published", "source": "fkie_nvd \u00b7 FKIE_CVE-2026-33359", "record": "FKIE_CVE-2026-33359", "field": "published", "metric_role": "disclosure", "metric_basis": "recorded_event", "linked_record": true}, {"timestamp": "2026-05-11T17:39:26.604Z", "date_only": false, "value": "2026-05-11T17:39:26.604637Z", "kind": "sighting", "label": "Seen", "source": "https://bsky.app/profile/thehackerwire.bsky.social/post/3mllsx6dccq2t", "record": "CVE-2026-33359", "field": "creation_timestamp", "uuid": "f993574e-39ce-431a-9a69-27920e18541a", "content": "\ud83d\udfe0 CVE-2026-33359 - High (7.5)\n\nIn Meari IoT Cloud alert image storage on Alibaba OSS (latest observed; storage service version n...\n\nhttps://www.thehackerwire.com/vulnerability/CVE-2026-33359/\n\n#infosec #cybersecurity #CVE #vulnerability #security #patchstack", "url": "https://bsky.app/profile/thehackerwire.bsky.social/post/3mllsx6dccq2t"}, {"timestamp": "2026-05-11T18:17:55.620Z", "date_only": false, "value": "2026-05-11T18:17:55.620113Z", "kind": "other", "label": "Recorded", "source": "cvelistv5 / CISA-ADP", "record": "CVE-2026-33359", "field": "containers.adp.0.metrics.0.other.content.timestamp"}, {"timestamp": "2026-05-11T18:17:55.620Z", "date_only": false, "value": "2026-05-11T18:17:55.620113Z", "kind": "other", "label": "Recorded", "source": "fkie_nvd \u00b7 FKIE_CVE-2026-33359", "record": "FKIE_CVE-2026-33359", "field": "metrics.ssvcV203.0.ssvcData.timestamp", "linked_record": true}, {"timestamp": "2026-05-11T18:18:06.184Z", "date_only": false, "value": "2026-05-11T18:18:06.184Z", "kind": "other", "label": "Updated", "source": "cvelistv5", "record": "CVE-2026-33359", "field": "cveMetadata.dateUpdated"}, {"timestamp": "2026-05-11T18:18:06.184Z", "date_only": false, "value": "2026-05-11T18:18:06.184Z", "kind": "other", "label": "Updated", "source": "cvelistv5 / CISA-ADP", "record": "CVE-2026-33359", "field": "containers.adp.0.providerMetadata.dateUpdated"}, {"timestamp": "2026-05-11T18:31:45Z", "date_only": false, "value": "2026-05-11T18:31:45Z", "kind": "milestone", "label": "Published", "source": "github \u00b7 GHSA-FXWQ-QQHR-34Q4", "record": "GHSA-FXWQ-QQHR-34Q4", "field": "published", "metric_role": "disclosure", "metric_basis": "recorded_event", "linked_record": true}, {"timestamp": "2026-05-11T18:31:45Z", "date_only": false, "value": "2026-05-11T18:31:45Z", "kind": "other", "label": "Updated", "source": "github \u00b7 GHSA-FXWQ-QQHR-34Q4", "record": "GHSA-FXWQ-QQHR-34Q4", "field": "modified", "linked_record": true}, {"timestamp": "2026-10-26T18:10:00.173Z", "date_only": false, "value": "2026-10-26T18:10:00.173", "kind": "other", "label": "Updated", "source": "fkie_nvd \u00b7 FKIE_CVE-2026-33359", "record": "FKIE_CVE-2026-33359", "field": "lastModified", "linked_record": true}, {"timestamp": "2026-10-31T14:00:29Z", "date_only": false, "value": "2026-10-31T14:00:29.000000Z", "kind": "sighting", "label": "Seen", "source": "https://gist.github.com/tartinesKiller/d3fe904153d95dc19bda9ee324358674", "record": "CVE-2026-33359", "field": "creation_timestamp", "uuid": "8fd59b81-1b1d-4cd9-bfa1-a0f1209a2936", "content": "# Keeping a CloudEdge/Meari white-label camera local: a WAN-minimal firewall recipe\n\n**Goal:** keep local ONVIF/RTSP streaming working (Blue Iris / Frigate / any NVR) while the camera is denied almost all Internet access \u2014 no P2P relay, no cloud uploads, no MQTT.\n\n**Tested with:** Boifun Baby 6T monitor (Amazon ASIN B0DT3ZJ7FH, app **BoifunCam**, backend **CloudEdge / Meari**, EU region), camera firmware **V5.5.4** (as shipped in 2026), OpenWrt firewall. The exact endpoints below were observed on *my* unit \u2014 yours may differ by region, brand or firmware. **The method is the durable part; the IPs are just an example.**\n\n**Why do this:** the Meari/CloudEdge platform was audited in 2026 ([nobody-puts-baby-in-a-corner](https://github.com/xn0tsa/nobody-puts-baby-in-a-corner)): 5 CVEs, including a P2P/static-password scheme and a device-status API that returns the WAN IP + relay parameters of any device by serial number (CVE-2026-33357). All flaws are platform-side; no firmware update on your camera will fix them. The audit's advice to owners is essentially to unplug the camera. This gist documents a middle ground that worked for one unit: *keep the camera, but reduce its Internet access to the minimum that preserves local streaming*.\n\n---\n\n## TL;DR\n\nThe firmware keeps its local ONVIF/RTSP services **closed until it considers itself \"online\" in the cloud**. That's why blocking the whole WAN kills local streaming even though the camera still answers ping. But being \"online\" only requires **3 outbound channels**. Everything else \u2014 the P2P relay used by the mobile app, cloud uploads, MQTT \u2014 can stay blocked:\n\n```\nALLOW (outbound, from camera):\n  1. API      apis-eu-oldroot.cloudedge360.com  \u2192 63.188.173.207:443/TCP   (device registration)\n  2. Keepalive euce.mearicloud.com             \u2192 47.91.73.19:9266/TCP     (presence)\n  3. Broker   8.211.0.0/16:18665/TCP                        (registration; IPs rotate!)\nDROP: everything else (zone policy: reject forward to WAN)\n```\n\nConsequences on my unit: local ONVIF/RTSP works indefinitely (Blue Iris 24/7), the vendor app can no longer stream at all (by design \u2014 see \"Trade-offs\" below), and measured egress stayed at control-plane levels (a few KB per minute, no media).\n\n---\n\n## 1. The core behavior to understand\n\n- Full WAN block \u2192 camera answers ping, but ports 8000 (ONVIF) and 554 (RTSP) are **closed**. The device only opens local services once its cloud registration succeeds. Same behavior reported by other CloudEdge users on ipcamtalk.\n- The mobile app streams **through the WAN P2P relay even when the phone is on the same LAN**. Measured: a local app session pushed \\~6.7 MB through `172.233.253.55:9100/UDP` in a few minutes. So an app that \"works at home\" is not proof that the video stays local \u2014 on this unit, it didn't.\n- The relay is **not needed** for local ONVIF/RTSP. It is only needed by the app's remote-streaming path.\n\n## 2. The method (works for any Meari white-label)\n\n### 2.1 Map what the camera talks to\n\nPut the camera in its own firewall zone / VLAN, default **reject** forward to WAN, then temporarily allow the traffic you see it attempting (or capture while everything is rejected). Capture SYNs at boot:\n\n```sh\n# OpenWrt \u2014 note: GNU timeout is prefixed on OpenWrt\ncoreutils-timeout 150 tcpdump -i any -nn -s0 -w /tmp/cam-boot.pcap \\\n  'src host 192.168.226.220 and (tcp[tcpflags] &amp; tcp-syn != 0 or udp port 53)'\n\n# Summarize attempted destinations (LINUX_SLL2: destination = field $7)\ntcpdump -nn -r /tmp/cam-boot.pcap 'tcp[tcpflags] &amp; tcp-syn != 0' \\\n  | awk '{print $7}' | sed 's/:$//' | sort | uniq -c | sort -rn | head -20\n\n# DNS queries in the clear (identifies the domains it resolves)\ntcpdump -nn -r /tmp/cam-boot.pcap 'udp port 53' | head -30\n```\n\nThe DNS capture is particularly useful: on my unit, only **two hostnames** are ever resolved \u2014 the rest of the endpoints (broker, relay, keepalive) are **IP addresses distributed by the protocol**, never seen in DNS. That distinction drives what kind of firewall rule you can write (see \u00a73).\n\n### 2.2 One change per iteration, always reboot\n\nFirewall keepalive behavior only matters at boot. The process I used:\n\n1. Remove (or disable) **one** allow rule.\n2. `fw4 reload`.\n3. Power-cycle the camera, wait \\~2 min.\n4. Test the local stream (`Test-NetConnection  -Port 8000` from Windows, or the NVR directly).\n5. Record the result, then iterate.\n\n### 2.3 What I ended up with (Boifun 6T, EU)\n\n\n| Decision | Endpoint                                                                   | Role                                                              | Result                                                                                                                                        |\n| -------- | -------------------------------------------------------------------------- | ----------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------- |\n| **KEEP** | `63.188.173.207:443/TCP` (`apis-eu-oldroot.cloudedge360.com`)              | API: registration, hands the camera its broker/relay addresses    | Required \u2014 removing it kills local ONVIF                                                                                                      |\n| **KEEP** | `47.91.73.19:9266/TCP` (`euce.mearicloud.com`)                             | Persistent presence/keepalive connection                          | Required \u2014 removing it kills local ONVIF                                                                                                      |\n| **KEEP** | `8.211.0.0/16:18665/TCP` (Alibaba broker)                                  | Registration channel; it is what makes the firmware feel \"online\" | Required \u2014 removing it kills local ONVIF                                                                                                      |\n| DROP     | P2P relay pool `172.232.0.0/13` (`TCP 9371` + `UDP 9100`)                  | Remote/local app streaming, two-way audio, remote PTZ             | App can no longer stream; local ONVIF unaffected. This is the endpoint exposed by CVE-2026-33357, so it is the main one worth keeping blocked |\n| DROP     | `3.124.20.170:443/TCP` (AWS ELB `meari-eu-oldroot-slb-\u2026elb.amazonaws.com`) | Platform backend                                                  | No observable effect                                                                                                                          |\n| DROP     | `47.91.73.19:443/TCP`                                                      | HTTPS facade of the same server                                   | No observable effect                                                                                                                          |\n| DROP     | `meari-*.oss-*.aliyuncs.com`                                               | Alert-snapshot uploads (see CVE-2026-33359)                       | No images leave the house. Keep blocked, always                                                                                               |\n| DROP     | `mqtts-*.meari.com.cn:8883`                                                | MQTT keepalive/alerts (CVE-2026-33356)                            | Never contacted by this firmware \u2014 blocked by default                                                                                         |\n\n\n*Note on the OSS row:* despite the `eu`/`oss-eu` in the bucket names (`meari-{eu,oss-eu,\u2026}.oss-{region}.aliyuncs.com`), the provider is **Alibaba Cloud OSS** in every case \u2014 the label refers to the storage region, not to a non-Alibaba host. Per the audit (CVE-2026-33359), alert JPEGs are uploaded there without authentication and never expire. On my unit, OSS was blocked from the start and no upload attempt appeared in my captures (uploads are likely alert-triggered rather than at boot), so this destination is verified from the audit rather than from the wire.\n\n## 3. Rotation: what can and cannot be pinned\n\n- **The broker rotates IPs at every boot.** Observed in `8.211.x.x` alone: `8.211.1.7`, `.23.241`, `.7.83`, `.22.157`, `.15.111` across five boots. A static-IP allowlist is hopeless \u2014 allow `8.211.0.0/16` on port `18665/TCP` only, or solve it with a DNS-fed nftset if your broker hostname is resolvable (mine isn't).\n- **The relay is a pool** (six IPs observed in `172.232.0.0/13`). If you *must* keep the app working, a port-based rule is the only thing that follows the rotation \u2014 but then you've reopened the P2P door. On a baby monitor, I chose not to.\n- **The 443 APIs are DNS-resolved at every boot** \u2014 those IPs can rotate too (AWS ELBs especially). If local streaming breaks one day without an obvious cause, suspect rotation first: re-run the boot capture and adjust the rule.\n- One mistake that is easy to make: the relay stream uses **`UDP 9100`**, so an allow rule written for TCP matches nothing. The rule still looks correct at first glance, which makes the resulting failure confusing to diagnose. Worth double-checking the protocol column in the capture before concluding anything.\n\n## 4. Proving nothing leaves the network\n\nMeasure rather than assume. With the minimal set live, capture 10 minutes of egress and count bytes per destination:\n\n```sh\ncoreutils-timeout 600 tcpdump -i any -nn -s0 -w /tmp/measure.pcap \\\n  'src host 192.168.226.220 and not dst net 192.168.0.0/16'\n\ntcpdump -nn -r /tmp/measure.pcap \\\n  | awk '{split($7,a,\":\"); b[a[1]]+=$NF; p[a[1]]++} END {for (d in b) printf \"%10d bytes %6d pkts  %s\\n\", b[d], p[d], d}' | sort -rn\n```\n\nInterpretation: control-plane chatter is **tens of KB over 10 minutes**. Video \u2014 any video \u2014 is megabytes per minute. If the biggest number you see is a few KB/s, nothing is streaming out.\n\n## 5. OpenWrt reference config (UCI)\n\nZone `cam` (camera VLAN) with forward-to-WAN policy `reject`, plus three accepts:\n\n```sh\nuci add firewall rule\nuci set firewall.@rule[-1].name='cam-api'\nuci set firewall.@rule[-1].src='cam'\nuci set firewall.@rule[-1].src_ip='192.168.226.220'\nuci set firewall.@rule[-1].dest='wan'\nuci set firewall.@rule[-1].dest_ip='63.188.173.207'\nuci set firewall.@rule[-1].proto='tcp'\nuci set firewall.@rule[-1].dest_port='443'\nuci set firewall.@rule[-1].target='ACCEPT'\n# \u2026repeat for 47.91.73.19:9266 and 8.211.0.0/16:18665\u2026\nuci commit firewall &amp;&amp; fw4 reload\n```\n\nNote the broker rule uses `dest_ip='8.211.0.0/16'` \u2014 CIDR ranges are accepted.\n\n## 6. NVR integration notes (Boifun / CloudEdge firmware)\n\n- ONVIF service runs on **port 8000**\n- ONVIF user is **`admin`** (fixed), password = the one you set in the app under *Settings \u2192 ONVIF*.\n- Works in Blue Iris (audio + PTZ) and as `onvif://admin:@:8000/onvif/device_service` for go2rtc/Frigate.\n\n## 7. Trade-offs (accepted, on purpose)\n\n- The vendor app no longer streams, remote or on the LAN. Given the platform's documented issues, this is the goal.\n- No more push notification from vendor's app.\n- If your unit has a dedicated screen (Boifun 6T does): the FHSS parent unit is fully independent of Wi-Fi/WAN and is unaffected by all of this.\n\n## 8. Caveats\n\n- Endpoints above are from **one EU unit, firmware V5.5.4**. Meari ships per-region stacks (`apis-eu-oldroot\u2026`, `openapi-euce\u2026`, `meari-us\u2026`); expect different IPs elsewhere. Map your own, minimize your own. If your camera runs a different firmware version, treat the endpoint list as unverified and re-run the capture.\n- A firmware update can change all of it. The capture-and-iterate method in \u00a72 is the part that survives.\n- Stale conntrack entries: `conntrack` keeps showing an `ESTABLISHED` entry up to \\~2h after a rule change, even when no traffic flows anymore. Check rule counters before concluding a blocked endpoint is still in use.\n\n## References\n\n- Security audit: [xn0tsa/nobody-puts-baby-in-a-corner](https://github.com/xn0tsa/nobody-puts-baby-in-a-corner) (5 CVEs, May 2026)\n- Related discussion: [ipcamtalk \u2014 how to get a camera to work without internet?](https://ipcamtalk.com/threads/how-to-get-a-camera-to-work-without-internet.67179/)\n- Generic VLAN/firewall segmentation guide: [VueVille \u2014 How to Block IP Cameras from Accessing the Internet](https://www.vueville.com/home-security/cctv/ip-cameras/how-to-block-ip-cameras-from-accessing-the-internet/)\n\n*Feel free to fork, translate or repost. This investigation was carried out with the assistance of Mistral AI. If this was useful, consider sharing your own results in the threads above \u2014 corrections and reports from other firmware versions are welcome.*", "url": "https://gist.github.com/tartinesKiller/d3fe904153d95dc19bda9ee324358674"}], "metrics": {"as_of": "2026-11-22T13:21:27.696Z", "anchors": {"origin": {"timestamp": "2026-05-11T16:00:00Z", "date_only": false, "value": "2026-05-11T16:00:00.000Z", "kind": "other", "label": "Made public", "source": "cvelistv5 / runZero", "record": "CVE-2026-33359", "field": "containers.cna.datePublic", "metric_role": "disclosure", "metric_basis": "recorded_event"}, "discovery": null, "report": null, "acknowledgement": null, "disclosure": {"timestamp": "2026-05-11T16:00:00Z", "date_only": false, "value": "2026-05-11T16:00:00.000Z", "kind": "other", "label": "Made public", "source": "cvelistv5 / runZero", "record": "CVE-2026-33359", "field": "containers.cna.datePublic", "metric_role": "disclosure", "metric_basis": "recorded_event"}, "exploit_publication": null, "exploitation": null, "patch": null}, "vendor_response_time": {"available": false, "missing": ["start", "end"]}, "time_to_exploit": {"available": false, "missing": ["end"]}, "disclosure_to_exploit": {"available": false, "missing": ["end"]}, "disclosure_to_exploit_publication": {"available": false, "missing": ["end"]}, "time_to_patch": {"available": false, "missing": ["start", "end"]}, "total_time_to_patch": {"available": false, "missing": ["start", "end"]}, "disclosure_to_patch": {"available": false, "missing": ["end"]}, "time_unpatched": {"available": true, "seconds": 16838487.696009, "hours": 4677.357693335833, "days": 194.88990388899305, "min_seconds": 16838487.696009, "max_seconds": 16838487.696009, "approximate": false, "ongoing": true, "order_known": true, "batch_listing": false, "end_before_start": false, "start": {"timestamp": "2026-05-11T16:00:00Z", "date_only": false, "value": "2026-05-11T16:00:00.000Z", "kind": "other", "label": "Made public", "source": "cvelistv5 / runZero", "record": "CVE-2026-33359", "field": "containers.cna.datePublic", "metric_role": "disclosure", "metric_basis": "recorded_event"}, "end": {"timestamp": "2026-11-22T13:21:27.696Z", "date_only": false, "label": "As of", "source": "Calculation clock"}}, "exploitation_to_patch": {"available": false, "missing": ["start"]}, "exploit_publication_to_patch": {"available": false, "missing": ["start"]}}}
