CTIPilot
AI-generated · no human review · verify critical claims against the linked source. how it works →
← Back to Daily brief 2026-05-16
NOTABLECVE-2023-42793exploitedupdatedNATOB2research

SentinelOne "Living Off the Pipeline": attackers subverted TeamCity, GitLab automation, self-hosted runners and developer workstations in 2025

SentinelOne: attackers turn build servers, CI/CD runners and developer workstations into the intrusion path

Analysis

SentinelOne's post of 2026-05-15 describes how attackers in 2025 went after the infrastructure that builds and ships software, not only its dependencies (SentinelOne, 2026-05-15). In one case attackers exploited a vulnerable self-hosted TeamCity server (the post links JetBrains' advisory for CVE-2023-42793), created a benign-looking build configuration that a trusted build agent ran with SYSTEM privileges, deployed a backdoor into internal environments and stayed undetected for more than a year (SentinelOne, 2026-05-15). JetBrains fixed CVE-2023-42793, an unauthenticated remote-code-execution flaw, in TeamCity On-Premises 2023.05.4 (JetBrains, 2023-09-21) and warned that backdoors planted before the patch persist after it (JetBrains, 2023-10-18). In an intrusion from SentinelOne's annual threat report, a compromised GitLab service-account token was used to create projects with malicious Ansible playbooks that the organization's pipeline then executed automatically (SentinelOne, 2026-05-15).

The post's other examples target people and runners: Contagious Interview sent developers in cryptocurrency and blockchain fraudulent job offers leading to fake skill-assessment sites, where a fabricated error told them to run commands that silently installed malware with access to their SSH keys, tokens and credentials, and the Sha1-Hulud campaign registered infected systems as self-hosted GitHub runners under attacker-defined names and used workflows that ran code when someone commented in a repository discussion (SentinelOne, 2026-05-15). Malicious versions of popular build packages ran reconnaissance scripts during installation that harvested tokens and inspected environments, and phished maintainer accounts appended malicious code to legitimate package updates (SentinelOne, 2026-05-15).

Triage: compiling, downloading packages, invoking scripts and opening connections are normal build activity, so a malicious job under a valid service account on a trusted runner looks legitimate unless the account acts outside its role; SentinelOne's discriminators are who created the job, what changed, where the runner was registered and which secrets were accessed (SentinelOne, 2026-05-15).

Cited evidence

After gaining access, they created a benign-looking build configuration that was executed by a trusted build agent running with SYSTEM privileges.

The attacker used the token to create projects containing malicious Ansible playbooks, which were then automatically executed by the organization’s CI/CD pipeline.

Build runners should be treated as high-value systems, monitored for suspicious child processes, credential access attempts, reverse tunnels, and persistence mechanisms.

If a malicious job executes under a valid service account on a trusted runner, the behavior may appear legitimate unless the account performs actions that clearly deviate from its role.

SentinelOne 2026-05-15

The vulnerability may enable an unauthenticated attacker with HTTP(S) access to a TeamCity server to perform a remote code execution (RCE) attack and gain administrative control of the TeamCity server.

JetBrains 2023-09-21

Updates1

Correction

The monitoring priorities SentinelOne gives are new runner registrations and unusual runner names, new or modified pipelines created by service accounts, jobs interacting with unfamiliar repositories, workflows that establish network tunnels, and suspicious child processes, credential access attempts, reverse tunnels and persistence on build runners (SentinelOne, 2026-05-15); the post names no audit-event type, source-address criterion or specific binaries. It walks through examples rather than a three-case taxonomy, and the rogue self-hosted runners and discussion-comment workflow triggers belong to the Sha1-Hulud campaign (SentinelOne, 2026-05-15). The TeamCity case links CVE-2023-42793, an unauthenticated remote-code-execution flaw fixed in TeamCity On-Premises 2023.05.4 (JetBrains, 2023-09-21); JetBrains warns that backdoors planted before patching are likely to persist after it, so a server patched after early October 2023 needs a compromise assessment (JetBrains, 2023-10-18).

Sources3

Revision history

  1. Published 2026-05-16-5bc123a0
  2. Correction 2026-10-21T1148Z-audit

    SentinelOne's monitoring advice covers new runner registrations and unusual runner names, pipelines created or changed by service accounts, jobs touching unfamiliar repositories, tunnel-opening workflows and suspicious child processes, credential access, reverse tunnels and persistence on build runners; it names no audit-event type, source-address criterion or specific binaries. The post walks through examples rather than a taxonomy of three cases, and attributes rogue self-hosted runners and comment-triggered workflows to Sha1-Hulud. The TeamCity case concerns CVE-2023-42793, fixed in TeamCity On-Premises 2023.05.4, whose pre-patch backdoors JetBrains says persist after patching.

    Changed: affected_products body classification confidence cves evidence headline sectors sources sourcing_note summary techniques title

PROVENANCE

AI-generated · no human review · this permalink is the shareable record for the finding · verify operationally critical claims against the linked primary source.