Picture a developer starting Thursday. Coffee in hand, they open GitHub and see a pull request from a colleague they’ve worked with for years. The account is real. The name is familiar. The change looks small. Nothing about it feels wrong.
That is exactly the point.
This week, security researchers revealed a campaign they call wfx-loot. Starting on Tuesday, an attacker used stolen credentials for 4,837 GitHub accounts to quietly open probe pull requests across more than 61,000 repositories. In over 1,800 of them, the malicious code ran inside CI pipelines and walked away with secrets: cloud keys, tokens and credentials that unlock far more than source code.
There was no zero day. Not a clever exploit of GitHub itself. The attackers simply logged in as people who were already trusted and used valid user and bot tokens to do what those people were allowed to do.
The Uncomfortable Truth: Trust Is the New Attack Surface
For years, we have built security around a simple idea: keep the bad actors out. Firewalls, perimeter controls and vulnerability scanners all ask the same question: is this coming from somewhere it shouldn’t?
wfx-loot didn’t come from somewhere it shouldn’t. It came from inside the circle of trust.
That is what makes campaigns like this so dangerous and so hard to stop with traditional tools:
When the attacker looks exactly like your team, the question is no longer who got in. It is is this person behaving like themselves?
Why Behavior Is the Only Signal That Can’t Be Faked
A stolen password can be used perfectly. A stolen token can be presented flawlessly. What an attacker can’t easily copy is the rhythm of a real person’s work: the repositories they usually touch, the hours they keep, the volume of changes they make, the systems they normally reach.
When 4,837 accounts suddenly start opening near-identical pull requests across tens of thousands of repositories, the credentials look fine. The behavior does not.
This is where security has to evolve, from checking identity to understanding intent.
How Seceon Helps Teams See What Credentials Hide
The Seceon Open Threat Management (OTM) platform is built on a simple belief: real security comes from context. By bringing identity, cloud, network and application telemetry into one native data pipeline, Seceon can spot the moment trusted access starts doing untrusted things.
Know When a Trusted Identity Stops Acting Like Itself
aiSecurity UIDGuard360 continuously learns how each user, service account and automation token normally behaves. When a developer account suddenly starts touching repositories it has never touched, or a bot token begins acting at an unusual scale or speed, aiSecurity UIDGuard360 raises the alarm, even though the credentials themselves are perfectly valid.
Protect the Cloud Keys Attackers Are Really After
The real prize in wfx-loot wasn’t code. It was the cloud secrets sitting inside the CI pipelines. Seceon’s cloud security modules help close that gap. Cloud Entitlement Management (CIEM), an extension of aiSecurity UIDGuard360, shows exactly which identities hold which cloud permissions and highlights excessive or risky access. aiDataGuard, Seceon’s data security posture management (DSPM), finds sensitive data and exposed secrets across cloud environments. Cloud Workload Protection (CWPP), which extends the aiXDR-PMax agent to cloud workloads, and Kubernetes Security Posture Management (KSPM) protects the workloads and clusters those keys can unlock.
Catch Stolen Secrets the Moment They’re Used
A stolen key is only dangerous once someone uses it. aiSIEM, aiXDR and NDR correlate activity across identity, endpoint, cloud and network, using more than 4,000 machine learning models to flag a cloud key being used from an unfamiliar location, at an unusual time, or against resources it has never accessed before. Seceon detects threats in under five minutes on average, while reducing false positives by 95%, so teams can act on real signals instead of drowning in noise.
Respond Before the Damage Spreads
When the clock is ticking, humans need help. SOAR and SeraAI AutoSOC can automatically revoke a compromised token, disable a suspicious account, or isolate an affected workload, with automated response in under 90 seconds. That speed matters when attackers are moving across thousands of repositories at once.
Stay Ahead of the Next Campaign
aiTIP360 brings threat intelligence and brand protection into the same view, helping teams track emerging supply-chain campaigns, monitor leaked credentials and spot early signs that their developers or projects are being targeted.
What Every Engineering and Security Leader Should Do Now
You don’t need to wait for the next headline to act. A few steps make an immediate difference:
The Bottom Line
wfx-loot is a reminder that the most dangerous attackers may not look like attackers at all. They may look like your best engineer, your build bot, or your most trusted contributor.
The organizations that stay ahead won’t be the ones with the strongest walls. They’ll be the ones who understand their people, systems and access well enough to notice when something feels off, even when every credential checks out.
That is the kind of security Seceon is built to deliver: unified, behavioral and fast enough to matter.
Ready to see what your credentials might be hiding? Talk to Seceon today.