Repair Guide · Antminer · Control Board · Firmware · Pool Security
Antminer Control Board Firmware Hijack: Detect and Recover
A coordinated pool-redirect campaign showed how fast a compromised control board can move fleet hashrate without touching a single hashboard. This guide covers what happened, where the entry point likely was, and the on-site triage sequence that tells you whether a reflash is enough or the board has to come out.
- ~64,000
- Miners reported redirected in the Nov 2026 incident
- ~13 EH/s
- Hijacked hashrate at peak of the event
- ~100 PH/s
- Hijacked hashrate after the redirect collapsed
In November 2026, roughly 64,000 miners — about 13 EH/s — were redirected away from their operators' pools and pointed at OCEAN. The redirect was not a hashboard fault and not a PSU failure. It was a control-board-level change to pool configuration, pushed across a fleet in a way that pointed to a shared management path rather than individual machines being touched one by one.
The economics of the attack are the part most operators miss. The attacker did not collect the redirected revenue. Miners' income from the diverted period is not recoverable through the pool. OCEAN's hashrate briefly climbed to roughly 50 EH/s before falling back, and the hijacked hashrate dropped to about 100 PH/s once the redirect was cut. In other words, the damage was measured in lost uptime and misattributed shares, not in a clean theft the operator can claw back.
What the incident actually looked like on site
Operators did not see hardware alarms. Fans ran normally, temps held, hashboards reported expected hashrate. What changed was the pool URL, worker name, and sometimes the fallback pool entries in the control board's configuration. From the miner's point of view it was working. From the operator's point of view, shares stopped arriving at the expected account.
That signature matters for triage: if hashrate is normal but credited shares are not, suspect configuration before you suspect silicon.
Suspected entry points
Two threads run through the incident reporting. The first is Luxor Commander, a fleet management tool that talks to control boards over their local APIs. The second is CoreX Hosting, where pool settings were found changed without authorization. Both point at the same class of weakness: a management plane with credentials that reach many control boards, where a single compromised session or weak credential can rewrite pool config at scale.
There were also earlier, smaller reports of pool settings being altered without operator action — before the November event. That pattern is worth taking seriously. A pool URL that changes once and gets "fixed" by an operator who assumes someone on the team edited it is exactly how a foothold survives.
Treat fleet management credentials as miner credentials
Any tool that can push pool config to a control board is, functionally, root on that board. Scope its credentials per site, rotate them on a schedule, and never let one management account reach every fleet you operate.
On-site triage sequence
Work through this in order. Each step narrows whether you are looking at a configuration change, a persistent firmware modification, or dead storage.
Step 1
1. Confirm the symptom class
Check reported hashrate against pool-side credited hashrate for the same worker. Normal local hashrate with missing pool credit means configuration. Both low means hardware or network.
Step 2
2. Pull the pool configuration
Read pool URL, worker, and fallback entries from the web UI and from the API. Compare against your inventory record. A fallback pool you did not configure is a red flag even if the primary looks correct.
Step 3
3. Capture UART boot log
Connect to the control board debug header and capture the full boot. You want to see whether U-Boot verifies the image, whether the kernel panics, and whether any init script rewrites pool config after boot.
Step 4
4. Test persistence
Restore correct pool settings, reboot, and re-read the config. If the wrong pool returns, the change is baked into the image or an init hook. Stop here and plan a reflash.
Step 5
5. Rotate and isolate
Rotate pool credentials and any management-plane credentials that reached the board. Move the unit to an isolated VLAN until it is reflashed or replaced.
Reading the UART log
The boot log is the fastest way to separate a hijacked image from failing storage. What you are looking for:
u-boot: verifying signature... FAILED— the image is unsigned or has been modified. Reflash from a vendor-signed or verified build.Kernel panic - not syncingbefore userspace — the image is broken or the storage is failing. Do not assume hijack first.mmcblk0: errororjffs2: error— storage-level faults. Reflashing may not stick; plan for board replacement.- A clean boot into userspace followed by pool config reverting after a few seconds — an init script or cron job is rewriting configuration. This is the signature of a persistent redirect.
If the log is clean and the config stays correct across a reboot, you are probably dealing with a one-time unauthorized change through a management API. Rotate credentials, audit who else had access, and keep the unit under observation for a week.
Pool settings audit
Do this across the fleet, not just on the affected unit. A single board with a changed fallback pool tells you the management path was used, and every board behind that path is a candidate.
Primary pool URL
Matches inventory record
Matches, but changed recently without a ticket
Fallback pools
Empty or your own backups
Unknown pool or OCEAN entry you did not add
Worker name
Your naming convention
Truncated, generic, or blank worker
Config after reboot
Unchanged
Reverts to wrong pool
Management credentials
Scoped per site, rotated
Shared across all sites, never rotated
When to reflash, when to replace
Reflash
Clean UART boot, readable storage, config reverts after reboot, or an unsigned image detected. Flash a verified image for the exact control board revision — BHB42601 and BHB42701 images are not interchangeable, and Amlogic and Xilinx boards take different images entirely.
Replace the board
Kernel panic before userspace, mmcblk0 or jffs2 errors, failed U-Boot signature verification that persists after a known-good flash, or a board that will not accept a signed image. Storage that is failing will not hold a clean image.
Two flashing rules prevent most of the mess. First, never interrupt a write — a power loss, network drop, or pulled ribbon mid-flash turns a recoverable board into a brick. Second, never flash an image from a Telegram or Discord link. If you cannot verify where the image came from, you are installing the next hijack yourself.
Preventing the next one
- Scope fleet management credentials per site and rotate them. One account that reaches every control board is one credential away from a repeat.
- Record pool URL, worker, and fallback entries in your inventory, and alert on drift rather than waiting for a revenue gap.
- Keep control boards on a management VLAN with no outbound access except to your own pool endpoints and update servers.
- Verify firmware images against a known hash before flashing, and match the image to the control board revision printed on the board itself.
- After any redirect event, audit every board behind the same management path — not just the ones that showed up in the revenue report.
The November 2026 redirect did not break any miners. It moved their shares. That is the uncomfortable lesson: a control board compromise leaves you with a fleet that looks healthy, hashes at full rate, and pays someone else. Configuration auditing is not an optional extra on a large farm — it is the only thing that catches this class of incident before the pool statement does.
Frequently asked questions
Can a hijacked Antminer control board be cleaned without reflashing?
Rarely. If the pool URL was changed through an authenticated API or web session, restoring pool entries and rotating credentials may hold. If the change survives a reboot, a factory reset, and a credential rotation, treat the firmware as untrusted and reflash from a verified image.
Does a pool redirect damage the hashboards?
No. A pool redirect only changes where shares are submitted. Hashboards, PSUs, and thermal behavior are unaffected, which is why the incident showed up as a revenue and accounting problem rather than a hardware failure.
How do I know whether to reflash or replace the control board?
Reflash when UART shows a clean boot into a modified userspace and the eMMC or SD is readable. Replace when you see kernel panics, mmcblk0 or jffs2 errors, failed U-Boot signature verification, or a board that will not accept a signed image at all.
Repair & maintenance
Get Your Miners Back to Full Hashrate
Chip-level diagnostics, component-level repair, and burn-in validation — on-site across the USA, no shipping required.