Detections
Explore public detection logic contributed by the community across SIEM and rule languages.
2,686 detections
Filters
Last updated
All Time
Detection languages
1,342
389
264
219
197
Contributors
620
506
358
209
102
Categories
20,439
11,829
5,873
5,151
4,864
2,686
Platforms
1,168
638
452
382
359
Products / Services
593
445
403
264
229
MITRE Techniques
1,884
787
525
521
510
CVEs
11
5
5
5
3
IDS Classtypes
189
133
34
10
7
IDS Protocols
243
58
33
22
12
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. Multiple inspection attempts within a 15-minute window indicate automated exfiltration of valuable intellectual property such as custom model configurations, system prompts, and internal model specifications. This activity represents unauthorized data disclosure that could enable competitive intelligence gathering, model replication, or preparation for advanced attacks against the AI infrastructure.
Detects a sequential pattern where a browser process (Chrome, Edge, or Safari) accesses an AI chat platform's conversation API, followed within 5 minutes by a network connection to a known unauthorized exfiltration destination (api.pbapi.xyz). This pattern suggests the potential use of a browser-based tool or extension (e.g., Poper Blocker or similar) to intercept and exfiltrate AI chat history.
Detects outbound network traffic associated with the Azazel operation, including connections to identified C2 infrastructure, open directory/upload listeners, domain-specific staging servers, and exfiltration attempts to MEGA cloud storage endpoints.
Detects an anomalous volume of file downloads from SharePoint or OneDrive by a user on an unmanaged, non-compliant device, shortly following a suspicious or risky sign-in event. The rule establishes a baseline for individual user behavior over 30 days and alerts when the current hourly download activity significantly exceeds that baseline.
This rule detects potentially compromised accounts by correlating suspicious authentication activity (such as risky sign-ins, impossible travel, or malicious OAuth/Mail-rule configurations) with a subsequent spike in SharePoint or OneDrive file access, downloads, or external link sharing within a one-hour window. This behavior is indicative of an adversary performing cloud-based data exfiltration after gaining unauthorized access to an account.
Detects network connections or DNS resolutions to known anonymous file-sharing, cloud storage, and messaging-bot API endpoints (e.g., transfer.sh, Telegram Bot API, Discord CDN) commonly abused by adversaries for data staging and exfiltration. This rule specifically targets non-browser processes that attempt to communicate with these endpoints, helping identify potentially malicious exfiltration activities.
Detects outbound network traffic associated with the Azazel operation, including connections to identified C2 infrastructure, open directory/upload listeners, domain-specific staging servers, and exfiltration attempts to MEGA cloud storage endpoints.
Detects a sequential pattern where a browser process (Chrome, Edge, or Safari) accesses an AI chat platform's conversation API, followed within 5 minutes by a network connection to a known unauthorized exfiltration destination (api.pbapi.xyz). This pattern suggests the potential use of a browser-based tool or extension (e.g., Poper Blocker or similar) to intercept and exfiltrate AI chat history.
This rule detects anomalous activity associated with the Antino backdoor, which uses Microsoft OneDrive as a dead-drop command-and-control (C2) channel. It identifies specific file paths related to heartbeat check-ins, operator tool drops, and exfiltration retrieval, summarizing events by user and IP to highlight potential malicious persistent communication and data transfer patterns.
Detects network connection attempts by known AI agent or development-related runtime processes (python, node, etc.) to external endpoints that are not contained within a defined allowlist of LLM and infrastructure providers. This is intended to identify potential data exfiltration or unauthorized command and control communication originating from LLM-powered applications.
Detects a specific multi-stage exfiltration pattern characterized by initial reconnaissance/fingerprinting probes against sensitive service endpoints (/health, /docs, /openapi.json) followed by unauthorized access to artifact storage endpoints (/files or /file?name=) from the same device against the same host within a short timeframe.
Detects the initialization of an AWS CloudShell session followed by file transfer operations (PutFile or GetFile). This behavior may indicate an attacker using CloudShell to stage, exfiltrate, or move tools within the cloud environment, potentially bypassing endpoint detection by operating within the cloud service infrastructure.
Detects potential insider threats or departing employees by correlating abnormal bulk data downloads from cloud repositories (SharePoint/OneDrive/Box/etc.) with subsequent sensitive data access and exfiltration signals. It uses baseline behavioral profiling, requires a threshold of sensitive file interactions, and mandates high-confidence exfiltration indicators (removable media usage, suspicious external domains, or multiple combined destination events) to reduce noise. Archive/staging activity is used as an optional confidence booster.
Detects the execution of known cloud synchronization utilities like rclone, megasync, or megacmd, which are frequently leveraged by ransomware operators to exfiltrate data to public cloud storage providers. The rule monitors both local process execution with typical command-line arguments and subsequent outbound network connections to common cloud storage domains.
Detects outbound network connections to known SectopRAT C2 infrastructure, specifically targeting the hardcoded IP 98.142.252.140 or the non-standard port 15847 often used for encrypted exfiltration.
Detects the use of legitimate data transfer utilities (s5cmd, rclone) configured to exfiltrate sensitive document types (.doc, .pdf, .sql, etc.) to S3-compatible cloud storage. The rule correlates this activity with recent access to files containing sensitive credentials or configuration data, a pattern observed in double-extortion campaigns.
This rule detects anomalous activity involving cloud-based shell services (AWS CloudShell, Azure Cloud Shell, GCP Cloud Shell). It triggers on two main conditions: (1) First-time usage of these shell environments by a principal within the detection window, and (2) The execution of commands associated with data staging, ingress tool transfer, or exfiltration (e.g., curl, wget, storage copy commands).
This rule detects potential browser-based credential and session theft by monitoring for a process that accesses multiple sensitive web browser artifacts (cookies, local state, session storage) across several user profiles in a short timeframe. It correlates this activity with the subsequent creation of a compressed archive in a temporary directory followed by an outbound network connection, indicating a multi-stage exfiltration workflow.
Detects the creation or modification of Exchange mailbox rules using the New-InboxRule or Set-InboxRule cmdlets that perform automatic forwarding or redirection of email to external, non-trusted domains. This technique is commonly employed during account takeover events to facilitate covert data exfiltration.
Detects M365 Copilot information extraction jailbreak attacks that attempt to obtain sensitive, classified, or comprehensive data through various social engineering techniques including fictional entity impersonation, bulk data requests, and privacy bypass attempts. The detection searches exported eDiscovery prompt logs for extraction keywords like "transcendent," "tell me everything," "confidential," "dump," "extract," "reveal," and "bypass" in the Subject_Title field, categorizing each attempt by extraction type and assigning severity levels (CRITICAL for classified/proprietary data, HIGH for bulk extraction or privacy bypass). Prompts are further analyzed for compound risk patterns such as "Confidential+Extraction" or "Bulk_Request+Bypass," filtering out low-severity cases to surface the most dangerous attempts to exfiltrate sensitive organizational information through AI manipulation.
Detects anomalous high-volume API querying patterns against an AI model, characterized by an abnormally high number of distinct input hashes and query frequency compared to the population baseline for that model. This pattern is consistent with model extraction or stealing attacks aimed at reconstructing a proxy model.
Page 3 of 128