Ankit Mehta
@SecvynIndiaTrusted contributorCompletionist
7 followers9,573 downloads10,241 copies160 likes55,511 views
8,655 detections
Filters
Last updated
All Time
Detection languages
5,827
1,081
677
439
436
Categories
2,646
2,040
1,304
944
814
Platforms
5,999
954
874
706
593
Products / Services
2,800
1,372
1,362
725
707
MITRE Techniques
2,087
2,043
1,442
1,092
1,085
CVEs
17
16
15
11
11
IDS Classtypes
250
132
95
68
64
IDS Protocols
405
140
49
36
14
This rule detects repeated access attempts (3 or more within an hour) to critical credential storage processes or files, specifically lsass.exe, the Security Account Manager (SAM) database, or the NTDS.dit file. Such activity is indicative of credential theft attempts by an adversary.
This rule detects a high volume of outbound network connections (20 or more within an hour) from a single device and process to non-RFC1918 IP addresses on ports commonly associated with Remote Access Trojans (RATs) or Command and Control (C2) communication. The ports monitored include 443, 8443, 8080, 80, 4444, 5555, 6666, and 7777. This pattern can indicate a compromised host attempting to establish C2 with an external adversary.
This rule detects the creation of a new Windows service (EventID 7045) where the service executable path contains both 'System32' and 'svchost', and the service description is either empty or contains 'Network'. This combination of characteristics can indicate the installation of a malicious service, potentially a backdoor or trojan, attempting to masquerade as a legitimate system service.
This rule detects potential spear phishing campaigns by identifying a high volume of emails (10 or more within an hour) from external senders (not @company.com) that contain keywords like 'password', 'account', or 'security' in the subject line, and also include either an attachment or a URL. This combination of factors suggests a targeted social engineering attempt.
This rule detects potential bulk phishing email campaigns by identifying a high volume of emails originating from external senders (not @company.com) with suspicious subject lines containing keywords like 'action required', 'verify account', or 'confirm identity'. The rule triggers if at least 50 such emails are sent to 20 or more unique recipients within a one-hour window.
This rule detects emails where the sender's address contains keywords like 'paypal', 'amazon', or 'apple' but does not originate from their official domains (e.g., '@paypal.com'). This pattern is indicative of lookalike domain phishing attempts, where attackers try to impersonate legitimate brands to trick recipients.
This rule detects a high volume (5 or more within an hour) of emails from the same sender containing compressed attachments (.zip, .rar, .7z) and subjects indicating documents, invoices, or reports. This pattern can be indicative of phishing campaigns attempting to deliver malware or gather information.
This rule detects potential phishing attempts by identifying emails with URLs that originate from outside the internal company domain (@company.com). It specifically looks for emails where the subject line contains keywords like 'click', 'confirm', or 'update', and where multiple such emails (3 or more) are sent from the same external sender to the same recipient.
This rule detects multiple unauthorized access attempts to files identified as machine learning models, weights, or checkpoints. It specifically looks for Security Event ID 4656 (Auditing of object access) where the object name contains keywords like 'model', 'weights', or 'checkpoint', and the access mask indicates unauthorized access attempts (0x0001 for ReadData or ListDirectory, 0x0002 for WriteData or AddFile). The rule then summarizes these attempts by user, computer, and hour, triggering an alert if 5 or more attempts are made within an hour.
This rule detects potential data poisoning attacks targeting machine learning models by monitoring Windows Security Event ID 4688 (a process was created) for command-line activity containing keywords indicative of model training or dataset modification, combined with terms suggesting malicious intent like 'corrupted', 'poison', or 'injection'. It aims to identify attempts to manipulate or compromise ML models through their training data.
Page 818 of 866