Detection rules

The following analytic detects Zeek HTTP requests using methods outside a configurable allowlist of standard and commonly used HTTP methods. Unusual methods can indicate protocol abuse, exploitation attempts, or command-and-control traffic. Review the method, source, destination, and URI to assess the activity.

The following analytic detects SSH connections presenting client banners that do not match known, common SSH client software by leveraging Zeek ssh logs. Unusual or unknown SSH client banners may indicate custom tooling, offensive frameworks, tunneling utilities, or malware using SSH for command-and-control or lateral movement.

The following analytic detects HTTP POST requests containing heavily encoded data in the post body, specifically looking for sequences of hex-encoded bytes or base64-encoded strings. It leverages Zeek HTTP logs to identify POST requests where the body contains eight or more consecutive hex escape sequences or sixteen or more consecutive base64 character groups.

The following analytic detects the wermgr.exe process creating an alternate stream in the temp directory. It leverages Sysmon EventID 15 to identify these actions. This activity is significant because wermgr.exe is typically associated with error reporting, not file creation. Such activity is significant as it may indicate RoguePlanet malware, which creates an alternate stream in the temp directory to execute malicious code.

Detects DNS queries to definitionupdates.microsoft.com or the go.microsoft.com fwlink redirect used for WD update downloads, when the querying process is not a Windows system component. BlueHammer utilizes these definition updates as part of its exploit chain.

The following analytic identifies the execution of commonly used NirSoft utilities on Windows systems. This activity is significant for a SOC because NirSoft utilities, while legitimate, can be used by adversaries for malicious purposes like credential theft or system reconnaissance. If confirmed malicious, this activity could lead to unauthorized access, data exfiltration, or further system compromise.

The following analytic detects the creation of files associated with the NirSoft tool bundles on Windows endpoints. NirSoft is a well-known provider of free, portable utilities that can be used for various system and network tasks. However, threat actors often leverage these tools for malicious purposes, such as credential harvesting, network reconnaissance, and data exfiltration.

The following analytic detects "cldapi.dll" being loaded by a process not associated with legitimate cloud sync activity. The Windows Cloud Files API (cldapi.dll) is abused by several local privilege escalation exploits.

The following analytic detects the execution of AdvancedRun.exe, a tool with capabilities similar to remote administration programs like PsExec. It identifies the process by its name or original file name and flags common command-line arguments. This detection leverages data from Endpoint Detection and Response (EDR) agents, focusing on process and command-line telemetry. Monitoring this activity is crucial as AdvancedRun can be used for remote code execution and configuration-based automation.

The following analytic identifies processes with command-line parameters associated with web browser credential dumping tools, specifically targeting behaviors used by Remcos RAT malware. It leverages data from Endpoint Detection and Response (EDR) agents, focusing on command-line executions and specific file paths. This activity is significant as it indicates potential credential theft, a common tactic in broader cyber-espionage campaigns.

Detects potential prompt injection or jailbreak attempts against Ollama API endpoints by identifying requests with abnormally long response times. Attackers often craft complex, layered prompts designed to bypass AI safety controls, which typically result in extended processing times as the model attempts to parse and respond to these malicious inputs.

Detects Ollama server errors and failures during model loading operations that may indicate malicious model injection, path traversal attempts, or exploitation of model loading mechanisms to achieve remote code execution. Adversaries may attempt to load specially crafted malicious models or exploit vulnerabilities in the model loading process to execute arbitrary code on the server.

Detects data leakage and exfiltration attempts targeting Ollama model metadata and configuration endpoints. Adversaries repeatedly query /api/show, /api/tags, and /api/v1/models to systematically extract sensitive model information including architecture details, fine-tuning parameters, system paths, Modelfile configurations, and proprietary customizations.

Detects critical service crashes, fatal errors, and abnormal process terminations in Ollama that may indicate exploitation attempts, resource exhaustion attacks, malicious input triggering unhandled exceptions, or deliberate denial of service attacks designed to disrupt AI model availability and degrade system stability.

The following analytic identifies the execution of tools built by NirSoft by detecting specific command-line arguments such as "/stext" and "/scomma". It leverages data from Endpoint Detection and Response (EDR) agents, focusing on process names, parent processes, and command-line executions. This activity is significant because NirSoft tools, while legitimate, can be exploited by attackers for malicious purposes such as credential theft or system reconnaissance.

The following analytic identifies ICMP traffic to external IP addresses with total bytes (sum of bytes in and bytes out) greater than 1,000 bytes. It leverages the Network_Traffic data model to detect large ICMP packet that aren't blocked and are directed toward external networks. We use All_Traffic.bytes in the detection to capture variations in inbound versus outbound traffic sizes, as significant discrepancies or unusually large ICMP exchanges can indicate information smuggling, covert

The following analytic detects unauthorized container uploads to AWS Elastic Container Service (ECR) by monitoring AWS CloudTrail events. It identifies instances where a new container is uploaded by a user not previously recognized as authorized.

The following analytic identifies the provisioning of a new MFA device shortly after a password reset. It detects this activity by correlating Windows Event Log events for password changes (EventID 4723, 4724) with PingID logs indicating device pairing. This behavior is significant as it may indicate a social engineering attack where a threat actor impersonates a valid user to reset credentials and add a new MFA device.

Identifies high Ollama runner counts and large model buffer allocations as possible resource pressure indicators. Memory values vary with model size and hardware, and these operational logs do not establish malicious intent; validate against host capacity and workload baselines.

Detects API reconnaissance and endpoint scanning activity against Ollama servers by identifying sources probing multiple API endpoints within short timeframes, particularly when using HEAD requests or accessing diverse endpoint paths, which indicates systematic enumeration to map the API surface, discover hidden endpoints, or identify vulnerabilities before launching targeted attacks.

Detects high Ollama API request volumes from a single client in a short time window. A burst may indicate automated abuse or contribute to resource contention, but GIN access logs do not establish malicious intent or service impact; validate the client and workload against local usage baselines.

Identifies threat-listed domains that are currently parked and may be used for staging.

Identifies newly registered domains that are later observed in WhoisFreaks threat feeds.

Identifies high-confidence malware, phishing, and spam domains from WhoisFreaks threat feeds.

Identifies unusually large NRD ingest volume that may indicate bulk registration campaigns.

Identifies Silverfort incidents involving bursts of authentications from one host to many other hosts. This behavior may indicate reconnaissance to identify targets for lateral movement.

Identifies Silverfort incidents involving repeated attempts to authenticate to a user account. This behavior may indicate a brute-force attempt to gain unauthorized access.

Identifies Silverfort incidents involving denied MFA requests accompanied by additional risk indicators, such as a new host or impossible travel. This behavior may indicate a user rejecting an unauthorized authentication attempt.

Identifies Silverfort incidents involving unusually frequent access to CIFS servers. This behavior may indicate reconnaissance of shared resources before lateral movement.

Identifies Silverfort incidents involving a security log being cleared on a domain controller. Clearing this log may remove evidence of account activity and indicate an attempt to evade investigation.