Arnold Chan
@slazTrusted contributorCompletionist
2 followers2,099 downloads4,375 copies51 likes26,092 views
7,555 detections
Filters
Last updated
All Time
Detection languages
5,650
1,006
394
169
98
Categories
2,356
1,807
1,233
1,013
924
Platforms
4,191
1,622
1,250
676
676
Products / Services
2,820
1,003
991
720
485
MITRE Techniques
2,087
1,520
1,413
1,070
536
CVEs
38
33
24
22
21
IDS Classtypes
477
179
104
79
52
IDS Protocols
554
156
144
86
18
Detects phishing messages that deliberately mix writing systems/scripts/homoglyphs in the Subject line as ONE corroborating signal within a broader, multi-signal phishing pattern. Hardened against false positives: brand/product tokens (e.g. Microsoft, PayPal) are stripped before the Latin-script check so a lone brand mention in non-Latin correspondence doesn't count as mixed-script; the suspicious-URL signal requires a high-confidence credential/urgency term (not generic business vocabulary); sender risk requires the domain to be new to both the recipient and the organization; and at least three independent signal categories (not two) must corroborate the language-confusion signal before the rule fires.
Detects phishing messages that deliberately mix writing systems/scripts/homoglyphs in the Subject line as ONE corroborating signal within a broader, multi-signal phishing pattern. Hardened against false positives: brand/product tokens (e.g. Microsoft, PayPal) are stripped before the Latin-script check so a lone brand mention in non-Latin correspondence doesn't count as mixed-script; the suspicious-URL signal requires a high-confidence credential/urgency term (not generic business vocabulary); sender risk requires the domain to be new to both the recipient and the organization; and at least three independent signal categories (not two) must corroborate the language-confusion signal before the rule fires.
Detects file, process, and network activity associated with the TraderTraitor/KelpDAO malware campaign based on known hashes, filenames, paths, and C2 infrastructure.
Detects malicious AI assistant links delivered via spearphishing emails that contain pre-populated prompt injection query parameters. The rule correlates the clicking of a high-signal URL (containing parameters like prompt, system, or lengthy search queries with malicious keywords) originating from an email with a subsequent active session to the same AI service within 5 minutes, confirming potential weaponized AI assistant session manipulation. Covers T1566.002, T1204.001
Detects rapid, multi-category reconnaissance activity initiated by AI-agent binaries (e.g., Claude, Cursor, ChatGPT). The rule identifies sessions that perform four or more distinct discovery operations—spanning account, network service, system information, and network configuration discovery—within a short time window, indicating potentially malicious autonomous exploration by an AI agent. Covers T1046, T1087, T1083, T1016
Detects a multi-stage attack pattern against publicly exposed AI service endpoints. The detection correlates: (1) Enumeration of sensitive AI/MCP-related API endpoints (minimum 3 distinct hits) in a 10-minute window; (2) Subsequent POST requests to AI service endpoints within 15 minutes of the enumeration; and (3) A spike of 5 or more server errors (HTTP 5xx) within 10 minutes following the POST activity. This pattern suggests automated reconnaissance followed by attempted exploitation of exposed AI model interfaces or orchestration services. Covers T1595.002, T1190
Detects rapid, machine-speed, sequential attack-path adaptation by suspected AI agent binaries. The rule monitors for a correlation of four stages occurring within a tight 20-minute window: 1) Failed service connections across three or more distinct protocols (SMB, SSH, HTTP, RDP, WinRM), 2) Credential access attempts (Logon events or execution of known credential-dumping tools), and 3) A lateral movement pivot to a new, previously un-targeted internal destination. The rule specifically keys on the InitiatingProcessId and InitiatingProcessCreationTime to ensure the sequence is attributed to a single process instance, distinguishing it from human manual penetration testing activity. Covers T1190, T1210, T1078
Detects anomalous multi-agent chaining where one AI-enabled agent instance initiates a connection to an AI API and subsequently hands off information to a second, distinct agent instance, which then performs an suspicious downstream action such as spawning a shell or initiating an unauthorized external network connection. This behavior is intended to identify potential abuse of AI-coding assistant tools for automated execution chains.
Detects signs of the ClosedQuorum implant by correlating at least two redundant persistence mechanisms (Registry Run keys, Scheduled Tasks, WMI execution) on the same host within a 10-minute window. The rule focuses on artifacts originating from suspicious user-writable paths or unsigned binaries, while excluding known legitimate installation paths.
Detects signs of the ClosedQuorum implant by correlating at least two redundant persistence mechanisms (Registry Run keys, Scheduled Tasks, WMI execution) on the same host within a 10-minute window. The rule focuses on artifacts originating from suspicious user-writable paths or unsigned binaries, while excluding known legitimate installation paths.
Page 188 of 756