I run a crypto investing platform, Wealtii, and about a week ago a Hacker bot spent 105 minutes trying to break into it while I was asleep. I read the whole thing in the logs the next morning.
This post is that log, with one addition the Medium version of this story does not have: for every move the bot made, I explain in plain English what stopped it, and what you can check on any platform to see whether the same wall exists there. No architecture tour. Just attack, defence, and the question it leaves you with.
The result first, so nobody has to scroll for it. No account was taken over. No admin access was gained. No money moved. The cloud account's audit trail shows zero changes.
00:16 UTC: mass account creation
The move. Traffic from a VPN exit in Brazil and a rented cloud server starts creating accounts. Six in 21 minutes, four on throwaway email domains.
What held. Registration is rate limited, and email verification is required before an account does anything. The accounts existed, but each one was a shell with nothing in it.
What did not hold. Four throwaway domains got past my disposable-email filter. The filter has gaps, and I would rather say so than pretend the opening round was clean.
Your check. Try to sign up with a well-known throwaway inbox. If it sails through with no verification step, the platform is cheap to flood with fake accounts, and that makes every later defence weaker.
00:22 UTC: password spray
The move. Nine password guesses in twenty seconds, spread across accounts. None landed.
What held. Two things, in order. Repeated failed sign-ins lock an account temporarily, so the spray stalls fast. And on Wealtii a correct password is not enough on its own: every sign-in also needs a one-time code sent to the account's email, so even a lucky guess hits a second door.
Your check. Sign in to your platform and count the steps. If a password alone gets you to your balance, a password alone gets an attacker there too. A second code on every sign-in is the baseline, not a premium feature.
00:35 UTC: ninety minutes of hunting for staff
The move. The bot stops caring about ordinary users and starts probing for staff accounts, trying admin@, support@ and info@ addresses through the session-recovery flow, over and over, for almost an hour and a half.
What held. The rate limiter on recovery requests. It clamped down early and stayed down. The bot never found out whether those addresses even had accounts.
Why this is the most important entry in the log. One user account is one balance. One staff account is every balance. The fastest way to rob a platform is to become the platform, and the bot's time budget shows it knew that.
Your check. Ask support, in writing, how admin access is granted and whether a stolen staff login could move customer funds. A specific answer is a good sign. "We take security seriously" is not an answer.
01:23 UTC: a machine-made identity
The move. A new account on a public throwaway inbox, with a Unix timestamp baked into the email address, generated three seconds before sign-up.
What held. Nothing needed to. This was the bot preparing a clean identity for what came next. The timestamp is simply proof that no human was at the keyboard.
Your check. None for you here, but it is worth knowing what automation looks like: identities minted to order, with no history and no reuse.
01:38 UTC: trying to become the admin by email
The move. The bot requests to change its account email to my own admin address, so that it would appear, to the system and to anyone reading, as the platform's administrator.
What held. An email change confirms through the new address, not the old one. The verification code went to the real admin inbox, which only I can read. You cannot claim an address you cannot open. After 101 seconds the bot cancelled.
Your check. Change your own email on the platform and watch where the code goes. If the new address is accepted without a code sent to it, anyone can claim any address, including the platform's own.
01:42 UTC: four fake identity documents
The move. Four files uploaded to the KYC verification system as identity documents.
What held. The upload pipeline checks file type and size and does not trust the file name. I later pulled the raw bytes and inspected them in isolation, never opening them as images: three identical blank one-pixel JPEGs and one flat single-colour square. Classic scanner probes, built to find a pipeline that will accept anything. Nothing hidden, nothing executable, nothing accepted as a document.
Your check. Before you upload a passport anywhere, ask three things: are documents checked by content rather than name, are they stored encrypted and out of public reach, and are they ever opened as anything other than an image. The bigger risk in KYC is not a fake upload. It is a platform that handles your real documents carelessly.
01:44 UTC: the rename to admin123
The move. The account changes its display name to "admin123".
What held. Names carry no permissions. Admin rights on Wealtii are a role assigned on the server, not a string a user can type. The rename was theatre.
Your check. Ask whether any permission on the platform is tied to a username, an email prefix, or a display name. The answer should be no, every time.
01:45 UTC: resetting its own password
The move. A password reset on the bot's own account, to map out how recovery works.
What held. Recovery goes through the account's verified email with a one-time code, and it is rate limited. Mapping it told the bot nothing it could use.
About 02:00 UTC: the code-injection payload
The move. The bot sets its display name to a script-injection payload and plants attack code in a support ticket. The code was written to run inside my admin panel the moment the ticket was opened, promote the attacker to admin, and retitle the page "PWNED".
What held. Three independent walls, and any one of them would have been enough:
- The backend stripped the code before saving the ticket. What reached the database was text with the dangerous parts removed.
- The admin panel renders user content as plain text, never as live code. I saw the attack printed on my screen as harmless characters.
- The admin endpoint the code tried to call does not exist. Even a perfect payload would have been shouting at a door that was never built. Your check. This is the one almost no investor asks and the one that decides whether a platform can be taken over through a text box. Ask: are support messages, display names and uploads cleaned before staff see them, and are they displayed as plain text? If the platform shows user content to staff as live content, every user is one paste away from being the admin.
02:00 UTC: six bait tickets, one marked URGENT
The move. Six support tickets filed in a burst, one flagged URGENT, to get a human to open the booby-trapped one quickly.
What held. The payload was already dead by the time any ticket could be opened, because of the three walls above. The urgency flag changed nothing, which is the point of designing so that opening a ticket is always safe.
Your check. Notice that the only human-targeted move in the whole log was an urgency flag. Urgency is the mechanism of social engineering, on a support desk and in your own inbox. Anything that pushes you to act fast deserves to be read slowly.
02:01 UTC: silence
After 105 minutes, the bot stopped. By the numbers: 2 attack origins, 7 fake accounts, 9 password guesses in 20 seconds, about 90 minutes of admin hunting, 1 email-hijack attempt, 4 fake IDs, 1 code-injection and admin-takeover payload, 6 bait tickets, 0 successful.
The ten-point check you can run on any platform
- Throwaway email test: can a disposable inbox register without verification?
- Sign-in count: does a password alone reach the balance?
- Withdrawal code: does moving money need a second code every time?
- Staff access: will support explain how admin rights are granted?
- Email change: does the code go to the new address?
- Document handling: checked by content, stored encrypted, never executed?
- Names and permissions: is anything tied to a username?
- Recovery flow: one-time code to the verified email, rate limited?
- User content: cleaned on save and shown to staff as plain text?
- Incident honesty: does the platform publish when it is attacked, and what it fixed? That last one is why this post exists.
What Wealtii does, and what this log does not prove
Every wall above is described in general terms on my security page: rate limits on sign-in, registration, verification and recovery; one-time codes on every sign-in and every withdrawal; type and size checks on uploads; input cleaning; one active session per account; throwaway email blocking. The log handed me two pieces of homework: the disposable-domain filter has gaps, and the security page does not yet describe how admin tools handle user content before staff see it. Both are being fixed.
What the log does not prove is that Wealtii is safe. One failed attack is one data point. There is no third-party security audit or penetration test yet, no deposit insurance, no investor-compensation scheme, and one person operating the platform. All of that is on the transparency page, and it belongs next to this story.
The current protections are on the Wealtii security page, and the list of what is still missing is on the transparency page. Read both, then run the ten checks on every platform you use.
Digital assets are volatile and you can lose some or all of the money you invest. No security measure removes all risk, and no platform should be called unhackable. This is educational content, not financial or security advice.



Top comments (1)
This is genuinely an eye-opening read! What I really appreciate is that you haven't just shared the attack story, but explained what happened, what worked, and even where the gaps still exist. That level of transparency matters, especially in the crypto space, where trust and security go hand in hand. The point about protecting admin access and treating every support ticket as a potential security risk really stood out to me. Also, the 10-point checklist is something every crypto investor should take seriously. Great work turning a real security incident into practical lessons for everyone!