Thread hijacking
Email thread hijacking is an attack where a malicious reply is inserted into a real conversation. The attacker writes from a compromised mailbox of one participant, or from a lookalike of it, and quotes the genuine history. The reply inherits the trust of the thread, then asks for a payment, a sign-in or a file to be opened.
Updated
How thread hijacking works
Thread hijacking works by borrowing a conversation that both sides already trust, then changing one detail in it.
- The attacker gets into one participant's mailbox, usually with a stolen password or session. See account takeover.
- The attacker reads quietly and looks for conversations about money, contracts or access.
- The attacker picks a live thread and waits for a natural moment: an invoice is due, a document is expected.
- The attacker replies inside the thread. The reply comes from the real mailbox, or from a lookalike domain with the stolen history pasted underneath.
- The reply changes one thing: new bank details, a new link to the document, a file to open.
- Rules set in the compromised mailbox hide the answers from its real owner, so nobody notices.
What a hijacked thread looks like
A hijacked thread looks like the next message in a conversation you were already having. Jordan Ellis has been arranging a payment with Megan Park at Northwind. This reply arrives on the same subject, with the real history quoted below it.
From: Megan Park <megan@northwind-supply.example.net>
To: Jordan Ellis <jordan@trotta.email>
Subject: Re: Invoice 4471 for October
Hi Jordan, one change before you pay. We moved banks this
week. Please use the new details in the attached PDF.
> On Tuesday, Megan Park <megan@northwind-supply.example> wrote:
> Invoice 4471 is attached, due on the 1st.The quoted message is genuine. The new one comes from a different domain. When the attacker is inside the real mailbox, even that difference is gone.
Signs of thread hijacking
The signs of thread hijacking are small changes inside a conversation that otherwise looks the same.
| Sign | What it means |
|---|---|
| A reply changes payment details, a link or a file mid-thread | The one change is the attack. Everything around it is real. |
| The sender's domain differs from earlier messages | The history was copied to a lookalike. |
| Recipients were quietly dropped, or Reply-To is new | The attacker is steering the conversation away from whoever would object. |
| The tone, sign-off or language shifts | A different writer has joined. |
| You ask for confirmation by reply and get a quick yes | You asked the attacker. Use the phone. |
Why thread hijacking gets through ordinary filtering
Thread hijacking gets through because a reply from a real compromised mailbox passes the technical checks. The domain is genuine, so SPF, DKIM and DMARC pass. The sender is a known correspondent, the subject matches an existing conversation, and there may be no link and no file.
Gmail and Outlook screen mail before delivery for spam and malware. A calm reply in a running thread gives a filter little to judge. Thread hijacking is one way business email compromise is carried out, and the FBI IC3 2025 report puts losses from that crime at $3.05 billion.
When an AI agent reads the mail
An AI agent trusts a thread more than a human does, because the thread is its context. Given a conversation, a model continues it. Quoted history reads as established fact, and the newest message reads as the current instruction.
An agent rarely compares the sender of the ninth message with the sender of the second. It replies to whoever the headers name, a new Reply-To included. And it acts at once: it can update a supplier record or send the requested file before any human looks.
How to stop thread hijacking
Thread hijacking is stopped by verifying the change outside the thread, and by keeping attackers out of mailboxes to begin with.
General controls
- Verify a change to payment details by phone, on a number you had before the request arrived. Do not verify by replying.
- Compare the sender's full address with the earliest messages in the thread.
- Require a second approver for new payees and changed bank details.
- Protect mailboxes with phishing-resistant sign-in, and review mailbox rules and forwarding. Hidden rules are how a hijack stays quiet.
At the door, with a pre-delivery inbox
A pre-delivery inbox screens every reply as a new message while the sending server is still connected, however familiar its thread. Trotta Mail checks the sender each time: SPF, DKIM and DMARC, lookalike domains and first contact. In the supplier payments use case, a message announcing changed bank details is held for a human before anyone pays.
For an agent, a thread's context comes as one block with boundaries around the mail and a trust level for the thread. Outgoing mail is screened too: a reply steered to someone new waits for a human. See concepts and the security model.
The limit is real. A well-written reply from a genuinely compromised mailbox may carry no technical tell and can be delivered. No filter catches everything, which is why the phone call stays in the process.
Questions
What is a reply chain attack?
A reply chain attack is another name for thread hijacking. The attacker replies inside an existing email conversation, using a compromised or imitated account, so the message arrives with the trust the conversation already earned.
Does DMARC stop thread hijacking?
No. A reply sent from the real, compromised mailbox passes SPF, DKIM and DMARC. DMARC helps when the attacker moves the thread to a domain that fails authentication.
What should I do if I think a thread was hijacked?
Stop acting on the thread and call the other party on a number you already had. Tell them their mailbox may be compromised. If money was sent, contact the bank at once and report it.