Machine Learning

An Elastic machine learning rule detects atypical user context switching on Windows systems, leveraging tools like 'runas,' which may indicate account takeover or privilege escalation, prompting defenders to investigate user accounts, activity timestamps, and source devices for potential compromise.

Elastic Security's machine learning rule `v3_windows_anomalous_process_creation_ea` detects unusual parent-child process relationships on Windows systems, indicating potential malware execution or persistence mechanisms and allowing for early detection of new or emerging threats that bypass traditional antivirus.

An Elastic Security machine learning rule detects rare and unusual processes on individual Windows hosts, indicating potential unauthorized services, malware execution, or persistence mechanisms.

An Elastic machine learning rule detects rare processes on Linux hosts, indicating potential persistence mechanisms, unauthorized services, or malware execution by an unknown threat actor, impacting system integrity and security.

An Elastic machine learning rule detects unusual Linux process discovery activity from atypical user contexts, indicating a potentially compromised account performing reconnaissance for privilege escalation or persistence on Linux systems.

Elastic has developed a machine learning detection rule to identify unusual user command activity related to system information discovery on Linux systems, indicating potential post-compromise reconnaissance for privilege escalation or persistence.

An Elastic machine learning rule detects anomalous access to the cloud instance metadata service by unusual Windows processes, indicating potential credential harvesting or sensitive data extraction by adversaries within cloud environments.

An Elastic machine learning rule detects rare and unusual DNS queries that indicate potential malicious network activity, including initial access via phishing or malicious documents, persistence, command-and-control (C2) communication, or data exfiltration attempts by adversaries.

Elastic's machine learning rule identifies rare and anomalous web user agents originating from local systems, indicating potential command-and-control, data exfiltration, or persistence activities by malware or specialized tools, enabling detection engineers to investigate unusual web browsing from non-browser processes.

Elastic's machine learning models detect clusters of suspicious Windows processes that share a common parent process and exhibit unusually high malicious probability scores, aiming to uncover stealthy attacks, including those leveraging Living off the Land Binaries (LOLBins) and masquerading techniques, which might otherwise evade traditional detection methods.

Elastic's machine learning job, utilizing the ProblemChild supervised model and unsupervised techniques, detects Windows hosts exhibiting clusters of suspicious processes with unusually high malicious probability scores, often indicative of defense evasion through Living Off The Land Binaries (LOLbins) and masquerading techniques.

A machine learning job from Elastic's ProblemChild integration detects suspicious Windows processes, classified as malicious by a supervised ML model and anomalous due to unusual user contexts identified by an unsupervised ML model, indicating potential misuse of LOLbins or masquerading tactics for defense evasion.

An Elastic machine learning rule detects an unusual spike in concurrent active Okta sessions initiated by a user, indicating potential adversary abuse of valid credentials for privilege escalation or persistence through the execution of multiple privileged operations.

Elastic's machine learning rule identifies anomalous execution of privileged commands by a user on Linux systems, indicative of potential privilege escalation or misuse of valid accounts.

An Elastic machine learning job has identified unusually high median command line entropy for privileged commands executed by a user on Linux, suggesting possible privileged access activity through obfuscated or complex command lines which can be a sign of suspicious or unauthorized use of privileged access, potentially indicating privilege escalation or defense evasion.

Elastic Security's machine learning rule detects a high count of source IP addresses establishing Remote Desktop Protocol (RDP) connections with a single destination IP, indicating potential lateral movement attempts by threat actors using multiple compromised systems for persistence and redundancy.

A machine learning detection rule identifies lateral movement by flagging an unusual spike in the number of destination IPs establishing Remote Desktop Protocol (RDP) connections with a single source IP, indicating an attacker attempting to expand access within the network to discover valuable assets or further access points.

An Elastic machine learning job detects unusually large file transfers by remote hosts, indicating potential lateral movement or data exfiltration by adversaries who consolidate data into single large files to avoid detection.

An Elastic machine learning rule detects potential Domain Generation Algorithm (DGA) activity, commonly used by malware for command and control (C2) communication, by analyzing DNS requests from source IP addresses to identify aggregate patterns indicative of DGA usage.

Elastic's Data Exfiltration Detection integration leverages machine learning to identify rare processes writing data to external devices, indicating potential data exfiltration by adversaries using benign-looking processes.

Elastic has released a machine learning detection rule designed to identify potential data exfiltration attempts by flagging anomalous spikes in the volume of data written to external devices, indicating illicit data copying or transfer activities by threat actors.

Elastic's machine learning job identifies potential data exfiltration activity to unusual geo-locations by detecting anomalies in network traffic patterns, indicating adversaries leveraging command and control channels to transfer data outside normal organizational patterns.

A machine learning job by Elastic detects potential data exfiltration by identifying anomalous network traffic patterns where high bytes are sent to an unusual destination port, suggesting data is being exfiltrated via command and control channels.

Elastic's machine learning rule detects potential data exfiltration by identifying anomalous network traffic, specifically large data transfers to unusual geo-locations via IP addresses, indicating possible exfiltration over command and control channels.

Elastic has developed a machine learning detection rule to identify anomalous compiler activity on Linux systems, which can indicate unauthorized software deployment, ad-hoc changes, or privilege escalation attempts by adversaries.

Elastic has released a machine learning detection rule designed to identify rare and unusual process executions across multiple Linux hosts within an entire fleet, aiming to uncover potential malware or suspicious behaviors indicative of persistence or other malicious activity.

This Elastic machine learning rule detects anomalous network activity originating from Linux processes that typically do not engage in network communication, signifying potential command-and-control, lateral movement, persistence, or data exfiltration activity, often via process exploitation or injection.

An Elastic machine learning rule detects unusual user logon times, which can indicate credential compromise or unauthorized access, particularly when attackers operate from different time zones or during non-business hours, prompting investigation into the affected user account and related activities.

Elastic's machine learning job detects rare PowerShell script executions on Windows hosts, identified by their script block hash, indicating potential malware activity or persistence mechanisms that deviate from an established baseline.

An Elastic machine learning job detects anomalous PowerShell script behavior, specifically focusing on unusual data characteristics like obfuscation, indicating potential malicious scripts adversaries use for execution and defense evasion on Windows systems.

Elastic Security's machine learning rule detects an unusual spike in Windows user account management events, including account creation, modification, or deletion, indicating potential privilege escalation or unauthorized activity by an adversary.

Elastic Security's statistical model identifies command-and-control (C2) beaconing activity in network logs on Windows and Linux systems by analyzing network traffic patterns and excluding known benign processes, enabling defenders to detect and respond to stealthy adversary communications for persistence and data exfiltration.

A high-severity vulnerability, CVE-2026-12480, affects the `keras-team/keras` library, enabling an arbitrary HDF5 file read via a virtual dataset bypass, potentially leading to sensitive information disclosure or exfiltration from systems utilizing the library.

A critical vulnerability, CVE-2025-71369, in `picklescan` versions prior to 0.0.28 allows remote attackers to bypass safety checks for malicious Python pickle files that utilize specific `torch.utils.data.datapipes` methods, enabling undetected embedded malicious code to execute during deserialization, which results in remote code execution (RCE) on the victim's system.

A critical deserialization vulnerability (CVE-2025-71356) in `picklescan` versions prior to 0.0.28 allows attackers to embed undetected malicious code within Python pickle files, leading to remote code execution when these files are loaded by victims.

CVE-2025-71345 describes a critical vulnerability in `picklescan` versions prior to 0.0.30, where attackers can embed undetected malicious code within pickle files that specifically invoke the `torch.utils.bottleneck.__main__.run_autograd_prof` function, leading to remote code execution upon deserialization by bypassing `picklescan`'s security checks.

An authentication bypass vulnerability in `backpropagate` versions >= 1.1.0 and < 1.2.0 allows unauthenticated attackers to gain full control over the Reflex web UI, even when HTTP Basic authentication is ostensibly enabled via the `--auth` flag, permitting data exfiltration, arbitrary training runs, HuggingFace Hub push, disk-fill DoS, and sensitive path discovery.

Stanza, an NLP library, is vulnerable to remote code execution (CVE-2026-54499) due to an unsafe fallback mechanism when loading PyTorch model files, allowing an attacker who can place a malicious pretrain or model file to achieve arbitrary code execution on systems processing NLP pipelines, leading to credential theft, backdoors, data exfiltration, and lateral movement.

A machine learning job has detected a suspicious Windows process, predicted to be malicious by the ProblemChild supervised ML model and found to be suspicious given its user context by an unsupervised ML model, indicating potential defense evasion activity involving LOLbins.

This rule detects unusual process spawned by a parent process, potentially indicating malicious activity involving LOLbins by leveraging machine learning to identify anomalous process creation patterns that evade conventional search rules.

A machine learning job detects unusual Windows processes, potentially Living off the Land binaries, on hosts not commonly associated with malicious activity, indicating possible defense evasion attempts.

A machine learning job has detected a spike in bytes written to an external device, which is anomalous and can signal illicit data copying or transfer activities, potentially leading to data exfiltration.

A machine learning job has detected potential data exfiltration activity to an unusual geographical region, specifically by region name, indicating exfiltration over command and control channels.

A machine learning job has detected an unusually high file size shared by a remote host, indicating potential lateral movement as attackers bundle data into a single large file transfer to evade detection when exfiltrating valuable information.

A machine learning job detected an unusually high mean of RDP session duration, indicative of potential lateral movement or persistent access attempts by adversaries abusing RDP.

A machine learning job detected a suspicious Windows process, predicted malicious by the ProblemChild model and flagged as an unusual child process name for its parent, potentially indicating LOLbins usage and evading traditional detection.

An Elastic machine learning job detects anomalous remote file transfers to unusual directories, indicating potential lateral movement by attackers attempting to bypass standard security monitoring.

The ProblemChild machine learning model has detected a user with suspicious Windows processes exhibiting unusually high malicious probability scores, potentially indicating defense evasion via masquerading or LOLbins.

A machine learning job detected potential DGA (domain generation algorithm) activity indicative of malware command and control (C2) channels, identifying source IP addresses making DNS requests with a high probability of being DGA-generated, a technique used by adversaries to evade detection.

A machine learning job has identified a user performing privileged operations in Okta from an uncommon source IP, indicating potential privileged access activity indicative of account compromise or privilege escalation.

A machine learning job has identified an unusually high median command line entropy for privileged commands executed by a user on Linux systems, suggesting possible privileged access activity through command lines, indicating potential obfuscation or unauthorized use of privileged access.

An Elastic machine learning rule detects unusual remote file transfers with rare extensions, potentially indicating lateral movement activity on a host and suggesting adversaries bypassing security measures.

A machine learning job has detected high bytes of data written to an external device, potentially indicating illicit data copying or transfer activities leading to data exfiltration over a physical medium such as USB.

A machine learning job has identified an unusual spike in Okta group privilege change events, indicating potential privileged access activity where attackers might be elevating privileges by adding themselves or compromised accounts to high-privilege groups, enabling further access or persistence.

A machine learning job detected a suspicious Windows process, predicted to be malicious by the ProblemChild supervised ML model and found to be unusual within the user's context, potentially indicating defense evasion techniques like masquerading or the use of LOLbins.

The ProblemChild machine learning model detected a rare Windows process indicative of defense evasion, potentially involving LOLbins, on a host not commonly associated with malicious activity.

A machine learning job detected a user performing privileged operations in Windows from an uncommon source IP, potentially indicating account compromise or privilege escalation.

A machine learning job detects a rare process writing data to an external device, potentially indicating data exfiltration masked by benign-looking processes.

A machine learning job detects an abnormal volume of remote file transfers, potentially indicating lateral movement by attackers attempting to blend in with normal network egress activity.

A machine learning job detected a user performing privileged operations in Okta from an uncommon device, potentially indicating a compromised account or insider threat attempting privilege escalation.