Loading vulnerability…
CVE-2020-11981 is a command injection vulnerability in Apache Airflow 1.10.10 and earlier affecting deployments that use CeleryExecutor. Airflow workers consume task messages from a Redis or RabbitMQ broker, and in the vulnerable implementation command data taken from broker messages could be executed without sufficient sanitization or validation. An attacker able to connect directly to the Celery broker can forge or inject a task message containing attacker-controlled command content, causing the Celery worker to execute arbitrary commands on the worker host. The flaw is rooted in over-trusting broker-supplied task data as safe execution input.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
Repository provides a working PoC for CVE-2020-11981 (Apache Airflow <=1.10.10) command injection/RCE when using CeleryExecutor with Redis as broker. The core exploit is `PoC/src/scanner.py` (Python): it connects to a specified Redis endpoint (`--targets ip:port`), crafts a Celery task message whose `task` is `airflow.executors.celery_executor.execute_command`, base64-encodes a body containing a command, and `LPUSH`es it into the Redis list/queue `default`. For verification, the payload command is hardcoded to `ping <listen_ip> -c 1`. The scanner simultaneously starts a Scapy sniffer on a chosen interface (auto-selects `eth0` or interface matching `--lip`) to capture ICMP echo requests; if it sees an ICMP request from the target IP after injection, it marks the target as vulnerable and writes results to JSON (`--output`). Structure: - `PoC/` contains a Dockerfile to run the scanner with Poetry dependencies (redis, scapy, psutil) and the PoC code under `PoC/src/`. - `Stand/` contains a docker-compose demo environment: Airflow 1.10.10 (custom image) configured for CeleryExecutor with Redis broker (no auth) and Postgres backend, plus Flower. Redis is exposed on 6379 and pinned to 172.18.0.2 on a custom bridge network. Overall purpose: demonstrate and semi-automate exploitation/verification of Airflow CeleryExecutor RCE by injecting malicious Celery tasks into an accessible Redis broker, using an out-of-band ICMP callback to confirm execution.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
3 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
An Apache Airflow Celery/Redis task-injection remote code execution issue referenced as historical lineage for forged broker messages causing worker execution.
A command injection vulnerability in Apache Airflow that allows execution of attacker-controlled commands on worker machines via unsanitized commands read from the message broker.
An Apache Airflow command injection issue in CeleryExecutor deployments that allows arbitrary command execution on workers if an attacker can access the broker.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.