Detections
Explore public detection logic contributed by the community across SIEM and rule languages.
258 detections
Filters
Last updated
All Time
Detection languages
117
32
27
22
16
Contributors
99
60
33
25
8
Categories
187
123
62
59
44
Platforms
39,951
7,276
6,526
3,865
3,641
258
Products / Services
146
48
47
28
22
MITRE Techniques
95
94
69
57
41
CVEs
10
5
5
5
5
IDS Classtypes
9
7
4
3
2
IDS Protocols
23
3
1
Detects HTTP requests carrying a Brotli Content-Encoding request body, which serves as the delivery vector for Node2Shell (CVE-2026-111497), an out-of-bounds write vulnerability in the Brotli decoder bundled with Node.js and Bun.
Detects the deployment of the ARTEX AI penetration testing framework via Docker/Docker-compose combined with the execution of offensive security and scanning tools (such as nmap, sqlmap, nuclei, httpx, apktool, frida) on the same host within a specified timeframe.
This rule detects the use of 'docker cp' or 'sbx cp' commands which are susceptible to the CVE-2026-17106 (CopyEscape) vulnerability. This vulnerability allows a compromised container to perform arbitrary file writes or overwrites on the host system during archive extraction, potentially leading to container escape and privilege escalation by replacing critical system binaries like runc.
The following analytic detects the execution of curl commands with insecure flags within the Cisco Isovalent environment. It identifies this activity by monitoring process execution logs for curl commands that use the -k or --insecure flags. This behavior is significant for a SOC as it could allow an attacker to bypass SSL/TLS verification, potentially exposing the Kubernetes infrastructure to man-in-the-middle attacks. If confirmed malicious, this activity could lead to data interception, service disruptions, or unauthorized access to sensitive information.
This detection identifies Docker containers that mount the host's root directory into the container filesystem.
Mounting the entire host root directory into a container effectively grants the container visibility and potential write access to all files on the host system.
If the container is running as root or with elevated capabilities (e.g., --privileged), the risk is significantly increased.
Mounting the entire host root directory into a container effectively grants the container visibility and potential write access to all files on the host system.
If the container is running as root or with elevated capabilities (e.g., --privileged), the risk is significantly increased.
This rule detects the use of 'docker cp' or 'sbx cp' commands which are susceptible to the CVE-2026-17106 (CopyEscape) vulnerability. This vulnerability allows a compromised container to perform arbitrary file writes or overwrites on the host system during archive extraction, potentially leading to container escape and privilege escalation by replacing critical system binaries like runc.
Detects container runtime processes (e.g., docker, podman, crictl) launching containers with elevated privilege flags (such as --privileged, --net=host, or mounting the host root directory) or pulling images from non-standard IP-based registries. These activities are indicative of potential container escapes, persistence mechanisms, or the use of untrusted/malicious container images.
Detects inbound HTTP POST requests to the /generate endpoint of a Hugging Face Text Generation Inference (TGI) server. Such requests indicate attempts to use or interact with the inference API, potentially without authentication, which could lead to unauthorized resource utilization or model interaction if the service is not properly secured.
Detects unauthorized creation or modification of AI agent system prompts, guardrails, or policy configuration files. This rule monitors for changes made by processes or users other than those identified as authorized CI/CD deployment agents or service accounts, helping to prevent unauthorized tampering with AI agent behavior or safety policies.
Detects adversarial backdoor triggers (such as pixel patches, watermark-like artifacts, or malicious token sequences) injected into AI inference inputs. The rule identifies these inputs by observing a combination of suspicious trigger patterns and significant, anomalous jumps in prediction confidence directed at a specific target class, indicative of a model integrity compromise.
Detects various techniques used to escape from Linux container isolation, including accessing container daemon control sockets (Docker/containerd), abusing namespace operations (nsenter/setns), mounting sensitive host filesystem paths (e.g., /proc/1/root), performing chroot escapes, and abusing privileged capabilities or socket-based execution.
Detects anomalous outbound network connections initiated by security and developer tooling (Trivy, Checkmarx, LiteLLM, Telnyx). This behavior is consistent with supply-chain attacks, such as those attributed to the UNC6780/TeamPCP threat group, where compromised tooling is leveraged to exfiltrate sensitive cloud credentials or API keys to external, non-vendor infrastructure.
Detects files, such as container image layers or extracted binaries, that contain a combination of CARBONATO botnet registry infrastructure artifacts and cryptocurrency wallet application strings. This indicates the presence of a trojanized wallet application originating from the CARBONATO operators' unauthenticated registry infrastructure.
Detects evidence of cryptocurrency mining (XMRig) and suspicious process masquerading on Linux systems, alongside container-related events referencing known malicious or suspicious strings such as 'GH0ST_C2' or 'fsociety'.
Detects unauthenticated POST requests to the TorchServe Management API '/models' endpoint containing a 'url=' parameter. This pattern is indicative of an attempt to perform remote model registration, which could be leveraged for Remote Code Execution (RCE) on the server hosting the TorchServe service.
This rule detects POST requests to the '/v2/repository/index' endpoint of a Triton Inference Server. This endpoint is part of the Model Repository control API and is typically used to manage or inspect models. Unauthorized access to this API can be indicative of reconnaissance or exploitation attempts against the inference server.
Detects unauthenticated POST requests directed at the '/v1/completions' endpoint of a vLLM server. This behavior is indicative of potential unauthorized access or exploitation attempts against an exposed LLM inference API that lacks authentication controls.
Detects unauthenticated HTTP POST requests targeting the '/v1/graphql' endpoint of a Weaviate vector database. This pattern is indicative of reconnaissance or exploitation attempts against exposed Weaviate API instances.
Detects suspicious process activity or file access patterns originating from within a container that indicate an attempt to escape to the host. This includes monitoring for the use of tools like nsenter, unshare, and mount, attempts to interact with the docker.sock, or attempts to access critical host paths like /proc/1/root or /host/etc, excluding events originating from authorized container management processes.
Detects interpreter processes (e.g., python, node, bash) typically used for AI agents accessing sensitive system files or directories outside of defined sandbox or workspace paths. This behavior suggests potential sandbox escape or unauthorized path traversal by an AI agent.
Detects unauthorized attempts to interact with the Docker API to create containers or pull images, specifically looking for patterns associated with the CARBONATO worm which utilizes the /containers/create endpoint with 'net-setup' parameters.
Page 1 of 13