A supervised machine learning model (ProblemChild) has identified a suspicious Windows process event with high probability of it being malicious activity. Alternatively, the model's blocklist identified the event as being malicious.

Read More

A supervised machine learning model (ProblemChild) has identified a suspicious Windows process event with low probability of it being malicious activity. Alternatively, the model's blocklist identified the event as being malicious.

Read More

Looks for compiler activity by a user context which does not normally run compilers. This can be the result of ad-hoc software changes or unauthorized software deployment. This can also be due to local privilege elevation via locally run exploits or malware activity.

Read More

Searches for rare processes running on multiple Linux hosts in an entire fleet or network. This reduces the detection of false positives since automated maintenance processes usually only run occasionally on a single machine but are common to all or many hosts in a fleet.

Read More

Searches for rare processes running on multiple hosts in an entire fleet or network. This reduces the detection of false positives since automated maintenance processes usually only run occasionally on a single machine but are common to all or many hosts in a fleet.

Read More

Identifies unusual parent-child process relationships that can indicate malware execution or persistence mechanisms. Malicious scripts often call on other applications and processes as part of their exploit payload. For example, when a malicious Office document runs scripts as part of an exploit payload, Excel or Word may start a script interpreter process, which, in turn, runs a script that downloads and executes malware. Another common scenario is Outlook running an unusual process when malware is downloaded in an email. Monitoring and identifying anomalous process relationships is a method of detecting new and emerging malware that is not yet recognized by anti-virus scanners.

Read More

A machine learning job detected unusually large numbers of DNS queries for a single top-level DNS domain, which is often used for DNS tunneling. DNS tunneling can be used for command-and-control, persistence, or data exfiltration activity. For example, dnscat tends to generate many DNS questions for a top-level domain as it uses the DNS protocol to tunnel data.

Read More

A machine learning job detected a rare destination country name in the network logs. This can be due to initial access, persistence, command-and-control, or exfiltration activity. For example, when a user clicks on a link in a phishing email or opens a malicious document, a request may be sent to download and run a payload from a server in a country which does not normally appear in network traffic or business work-flows. Malware instances and persistence mechanisms may communicate with command-and-control (C2) infrastructure in their country of origin, which may be an unusual destination country for the source network.

Read More

A machine learning job detected an unusual error in a CloudTrail message. These can be byproducts of attempted or successful persistence, privilege escalation, defense evasion, discovery, lateral movement, or collection.

Read More

A machine learning job found an unusual user name in the authentication logs. An unusual user name is one way of detecting credentialed access by means of a new or dormant user account. An inactive user account (because the user has left the organization) that becomes active may be due to credentialed access using a compromised account password. Threat actors will sometimes also create new users as a means of persisting in a compromised web application.

Read More

A machine learning job detected a significant spike in the rate of a particular error in the CloudTrail messages. Spikes in error messages may accompany attempts at privilege escalation, lateral movement, or discovery.

Read More

A machine learning job found an unusually large spike in authentication failure events. This can be due to password spraying, user enumeration or brute force activity and may be a precursor to account takeover or credentialed access.

Read More

A machine learning job detected an unusually large spike in network traffic that was denied by network access control lists (ACLs) or firewall rules. Such a burst of denied traffic is usually caused by either 1) a mis-configured application or firewall or 2) suspicious or malicious activity. Unsuccessful attempts at network transit, in order to connect to command-and-control (C2), or engage in data exfiltration, may produce a burst of failed connections. This could also be due to unusually large amounts of reconnaissance or enumeration traffic. Denial-of-service attacks or traffic floods may also produce such a surge in traffic.

Read More

A machine learning job found an unusually large spike in successful authentication events. This can be due to password spraying, user enumeration or brute force activity.

Read More

A machine learning job detected an unusually large spike in network traffic. Such a burst of traffic, if not caused by a surge in business activity, can be due to suspicious or malicious activity. Large-scale data exfiltration may produce a burst of network traffic; this could also be due to unusually large amounts of reconnaissance or enumeration traffic. Denial-of-service attacks or traffic floods may also produce such a surge in traffic.

Read More

A machine learning job detected an unusually large spike in network activity to one destination country in the network logs. This could be due to unusually large amounts of reconnaissance or enumeration traffic. Data exfiltration activity may also produce such a surge in traffic to a destination country that does not normally appear in network traffic or business workflows. Malware instances and persistence mechanisms may communicate with command-and-control (C2) infrastructure in their country of origin, which may be an unusual destination country for the source network.

Read More

A machine learning job found an unusually large spike in successful authentication events from a particular source IP address. This can be due to password spraying, user enumeration or brute force activity.

Read More

A machine learning job detected a PowerShell script with unusual data characteristics, such as obfuscation, that may be a characteristic of malicious PowerShell script text blocks.

Read More

A machine learning job detected an AWS API command that, while not inherently suspicious or abnormal, is being made by a user context that does not normally use the command. This can be the result of compromised credentials or keys as someone uses a valid account to persist, move laterally, or exfiltrate data.

Read More

A machine learning job detected AWS command activity that, while not inherently suspicious or abnormal, is sourcing from a geolocation (city) that is unusual for the command. This can be the result of compromised credentials or keys being used by a threat actor in a different geography than the authorized user(s).

Read More

A machine learning job detected AWS command activity that, while not inherently suspicious or abnormal, is sourcing from a geolocation (country) that is unusual for the command. This can be the result of compromised credentials or keys being used by a threat actor in a different geography than the authorized user(s).

Read More

A machine learning job detected a rare and unusual DNS query that indicate network activity with unusual DNS domains. This can be due to initial access, persistence, command-and-control, or exfiltration activity. For example, when a user clicks on a link in a phishing email or opens a malicious document, a request may be sent to download and run a payload from an uncommon domain. When malware is already running, it may send requests to an uncommon DNS domain the malware uses for command-and-control communication.

Read More

A machine learning job detected a user logging in at a time of day that is unusual for the user. This can be due to credentialed access via a compromised account when the user and the threat actor are in different time zones. In addition, unauthorized user activity often takes place during non-business hours.

Read More

Identifies Linux processes that do not usually use the network but have unexpected network activity, which can indicate command-and-control, lateral movement, persistence, or data exfiltration activity. A process with unusual network activity can denote process exploitation or injection, where the process is used to run persistence mechanisms that allow a malicious actor remote access or control of the host, data exfiltration, and execution of unauthorized network applications.

Read More

Looks for commands related to system network configuration discovery from an unusual user context. This can be due to uncommon troubleshooting activity or due to a compromised account. A compromised account may be used by a threat actor to engage in system network configuration discovery in order to increase their understanding of connected networks and hosts. This information may be used to shape follow-up behaviors such as lateral movement or additional discovery.

Read More

Looks for commands related to system network connection discovery from an unusual user context. This can be due to uncommon troubleshooting activity or due to a compromised account. A compromised account may be used by a threat actor to engage in system network connection discovery in order to increase their understanding of connected services and systems. This information may be used to shape follow-up behaviors such as lateral movement or additional discovery.

Read More

Identifies unusual destination port activity that can indicate command-and-control, persistence mechanism, or data exfiltration activity. Rarely used destination port activity is generally unusual in Linux fleets, and can indicate unauthorized access or threat actor activity.

Read More

Looks for anomalous access to the metadata service by an unusual process. The metadata service may be targeted in order to harvest credentials or user data scripts containing secrets.

Read More

Looks for commands related to system process discovery from an unusual user context. This can be due to uncommon troubleshooting activity or due to a compromised account. A compromised account may be used by a threat actor to engage in system process discovery in order to increase their understanding of software applications running on a target host or network. This may be a precursor to selection of a persistence mechanism or a method of privilege elevation.

Read More

Looks for commands related to system information discovery from an unusual user context. This can be due to uncommon troubleshooting activity or due to a compromised account. A compromised account may be used to engage in system information discovery in order to gather detailed information about system configuration and software versions. This may be a precursor to selection of a persistence mechanism or a method of privilege elevation.

Read More

Looks for anomalous access to the cloud platform metadata service by an unusual user. The metadata service may be targeted in order to harvest credentials or user data scripts containing secrets.

Read More

Looks for commands related to system user or owner discovery from an unusual user context. This can be due to uncommon troubleshooting activity or due to a compromised account. A compromised account may be used to engage in system owner or user discovery in order to identify currently active or primary users of a system. This may be a precursor to additional discovery, credential dumping or privilege elevation activity.

Read More

A machine learning job detected activity for a username that is not normally active, which can indicate unauthorized changes, activity by unauthorized users, lateral movement, or compromised credentials. In many organizations, new usernames are not often created apart from specific types of system activities, such as creating new accounts for new employees. These user accounts quickly become active and routine. Events from rarely used usernames can point to suspicious activity. Additionally, automated Linux fleets tend to see activity from rarely used usernames only when personnel log in to make authorized or unauthorized changes, or threat actors have acquired credentials and log in for malicious purposes. Unusual usernames can also indicate pivoting, where compromised credentials are used to try and move laterally from one host to another.

Read More

Identifies an unusually high number of authentication attempts.

Read More

A machine learning job detected an unusual network destination domain name. This can be due to initial access, persistence, command-and-control, or exfiltration activity. For example, when a user clicks on a link in a phishing email or opens a malicious document, a request may be sent to download and run a payload from an uncommon web server name. When malware is already running, it may send requests to an uncommon DNS domain the malware uses for command-and-control communication.

Read More

Identifies rare processes that do not usually run on individual hosts, which can indicate execution of unauthorized services, malware, or persistence mechanisms. Processes are considered rare when they only run occasionally as compared with other processes running on the host.

Read More

Identifies rare processes that do not usually run on individual hosts, which can indicate execution of unauthorized services, malware, or persistence mechanisms. Processes are considered rare when they only run occasionally as compared with other processes running on the host.

Read More

A machine learning job detected a user logging in from an IP address that is unusual for the user. This can be due to credentialed access via a compromised account when the user and the threat actor are in different locations. An unusual source IP address for a username could also be due to lateral movement when a compromised account is used to pivot between hosts.

Read More

Looks for sudo activity from an unusual user context. An unusual sudo user could be due to troubleshooting activity or it could be a sign of credentialed access via compromised accounts.

Read More

A machine learning job detected a rare and unusual URL that indicates unusual web browsing activity. This can be due to initial access, persistence, command-and-control, or exfiltration activity. For example, in a strategic web compromise or watering hole attack, when a trusted website is compromised to target a particular sector or organization, targeted users may receive emails with uncommon URLs for trusted websites. These URLs can be used to download and run a payload. When malware is already running, it may send requests to uncommon URLs on trusted websites the malware uses for command-and-control communication. When rare URLs are observed being requested for a local web server by a remote source, these can be due to web scanning, enumeration or attack traffic, or they can be due to bots and web scrapers which are part of common Internet background traffic.

Read More

A machine learning job detected a rare and unusual user agent indicating web browsing activity by an unusual process other than a web browser. This can be due to persistence, command-and-control, or exfiltration activity. Uncommon user agents coming from remote sources to local destinations are often the result of scanners, bots, and web scrapers, which are part of common Internet background traffic. Much of this is noise, but more targeted attacks on websites using tools like Burp or SQLmap can sometimes be discovered by spotting uncommon user agents. Uncommon user agents in traffic from local sources to remote destinations can be any number of things, including harmless programs like weather monitoring or stock-trading programs. However, uncommon user agents from local sources can also be due to malware or scanning activity.

Read More

Identifies Windows processes that do not usually use the network but have unexpected network activity, which can indicate command-and-control, lateral movement, persistence, or data exfiltration activity. A process with unusual network activity can denote process exploitation or injection, where the process is used to run persistence mechanisms that allow a malicious actor remote access or control of the host, data exfiltration, and execution of unauthorized network applications.

Read More

Identifies processes started from atypical folders in the file system, which might indicate malware execution or persistence mechanisms. In corporate Windows environments, software installation is centrally managed and it is unusual for programs to be executed from user or temporary directories. Processes executed from these locations can denote that a user downloaded software directly from the Internet or a malicious script or macro executed malware.

Read More

Looks for anomalous access to the metadata service by an unusual process. The metadata service may be targeted in order to harvest credentials or user data scripts containing secrets.

Read More

A machine learning job detected an unusual remote desktop protocol (RDP) username, which can indicate account takeover or credentialed persistence using compromised accounts. RDP attacks, such as BlueKeep, also tend to use unusual usernames.

Read More

A machine learning job detected an unusual Windows service, This can indicate execution of unauthorized services, malware, or persistence mechanisms. In corporate Windows environments, hosts do not generally run many rare or unique services. This job helps detect malware and persistence mechanisms that have been installed and run as a service.

Read More

Looks for anomalous access to the cloud platform metadata service by an unusual user. The metadata service may be targeted in order to harvest credentials or user data scripts containing secrets.

Read More

A machine learning job detected an unusual user context switch, using the runas command or similar techniques, which can indicate account takeover or privilege escalation using compromised accounts. Privilege elevation using tools like runas are more commonly used by domain and network administrators than by regular Windows users.

Read More

A machine learning job detected activity for a username that is not normally active, which can indicate unauthorized changes, activity by unauthorized users, lateral movement, or compromised credentials. In many organizations, new usernames are not often created apart from specific types of system activities, such as creating new accounts for new employees. These user accounts quickly become active and routine. Events from rarely used usernames can point to suspicious activity. Additionally, automated Linux fleets tend to see activity from rarely used usernames only when personnel log in to make authorized or unauthorized changes, or threat actors have acquired credentials and log in for malicious purposes. Unusual usernames can also indicate pivoting, where compromised credentials are used to try and move laterally from one host to another.

Read More

A supervised machine learning model has identified a DNS question name that is predicted to be the result of a Domain Generation Algorithm (DGA), which could indicate command and control network activity.

Read More

A supervised machine learning model has identified a DNS question name with a high probability of sourcing from a Domain Generation Algorithm (DGA), which could indicate command and control network activity.

Read More

A supervised machine learning model has identified a DNS question name that used by the SUNBURST malware and is predicted to be the result of a Domain Generation Algorithm.

Read More

A population analysis machine learning job detected potential DGA (domain generation algorithm) activity. Such activity is often used by malware command and control (C2) channels. This machine learning job looks for a source IP address making DNS requests that have an aggregate high probability of being DGA activity.

Read More

A machine learning job has detected unusually high number of process arguments in an RDP session. Executing sophisticated attacks such as lateral movement can involve the use of complex commands, obfuscation mechanisms, redirection and piping, which in turn increases the number of arguments in a command.

Read More

A machine learning job has detected unusually high mean of RDP session duration. Long RDP sessions can be used to evade detection mechanisms via session persistence, and might be used to perform tasks such as lateral movement, that might require uninterrupted access to a compromised machine.

Read More

A machine learning job has detected unusually high variance of RDP session duration. Long RDP sessions can be used to evade detection mechanisms via session persistence, and might be used to perform tasks such as lateral movement, that might require uninterrupted access to a compromised machine.

Read More

A machine learning job has detected data exfiltration to a particular destination port. Data transfer patterns that are outside the normal traffic patterns of an organization could indicate exfiltration over command and control channels.

Read More

A machine learning job has detected data exfiltration to a particular geo-location (by IP address). Data transfers to geo-locations that are outside the normal traffic patterns of an organization could indicate exfiltration over command and control channels.

Read More

A machine learning job has detected data exfiltration to a particular geo-location (by region name). Data transfers to geo-locations that are outside the normal traffic patterns of an organization could indicate exfiltration over command and control channels.

Read More

A machine learning job has detected data exfiltration to a particular geo-location (by region name). Data transfers to geo-locations that are outside the normal traffic patterns of an organization could indicate exfiltration over command and control channels.

Read More

A machine learning job has detected high bytes of data written to an external device. In a typical operational setting, there is usually a predictable pattern or a certain range of data that is written to external devices. An unusually large amount of data being written is anomalous and can signal illicit data copying or transfer activities.

Read More

A machine learning job has detected high bytes of data written to an external device via Airdrop. In a typical operational setting, there is usually a predictable pattern or a certain range of data that is written to external devices. An unusually large amount of data being written is anomalous and can signal illicit data copying or transfer activities.

Read More

A machine learning job has detected a high count of destination IPs establishing an RDP connection with a single source IP. Once an attacker has gained access to one system, they might attempt to access more in the network in search of valuable assets, data, or further access points.

Read More

A machine learning job has detected a high count of source IPs establishing an RDP connection with a single destination IP. Attackers might use multiple compromised systems to attack a target to ensure redundancy in case a source IP gets detected and blocked.

Read More

A machine learning job has detected unusually high number of processes started in a single RDP session. Executing a large number of processes remotely on other machines can be an indicator of lateral movement activity.

Read More

A machine learning job has detected an abnormal volume of remote files shared on the host indicating potential lateral movement activity. One of the primary goals of attackers after gaining access to a network is to locate and exfiltrate valuable information. Attackers might perform multiple small transfers to match normal egress activity in the network, to evade detection.

Read More

A machine learning job combination has detected a set of one or more suspicious Windows processes with unusually high scores for malicious probability. These process(es) have been classified as malicious in several ways. The process(es) were predicted to be malicious by the ProblemChild supervised ML model. If the anomaly contains a cluster of suspicious processes, each process has the same host name, and the aggregate score of the event cluster was calculated to be unusually high by an unsupervised ML model. Such a cluster often contains suspicious or malicious activity, possibly involving LOLbins, that may be resistant to detection using conventional search rules.

Read More

A machine learning job combination has detected a set of one or more suspicious Windows processes with unusually high scores for malicious probability. These process(es) have been classified as malicious in several ways. The process(es) were predicted to be malicious by the ProblemChild supervised ML model. If the anomaly contains a cluster of suspicious processes, each process has the same parent process name, and the aggregate score of the event cluster was calculated to be unusually high by an unsupervised ML model. Such a cluster often contains suspicious or malicious activity, possibly involving LOLbins, that may be resistant to detection using conventional search rules.

Read More

A machine learning job combination has detected a set of one or more suspicious Windows processes with unusually high scores for malicious probability. These process(es) have been classified as malicious in several ways. The process(es) were predicted to be malicious by the ProblemChild supervised ML model. If the anomaly contains a cluster of suspicious processes, each process has the same user name, and the aggregate score of the event cluster was calculated to be unusually high by an unsupervised ML model. Such a cluster often contains suspicious or malicious activity, possibly involving LOLbins, that may be resistant to detection using conventional search rules.

Read More

A machine learning job has detected a suspicious Windows process. This process has been classified as suspicious in two ways. It was predicted to be suspicious by the ProblemChild supervised ML model, and it was found to be an unusual process, on a host that does not commonly manifest malicious activity. Such a process may be an instance of suspicious or malicious activity, possibly involving LOLbins, that may be resistant to detection using conventional search rules.

Read More

A machine learning job has detected a suspicious Windows process. This process has been classified as malicious in two ways. It was predicted to be malicious by the ProblemChild supervised ML model, and it was found to be an unusual child process name, for the parent process, by an unsupervised ML model. Such a process may be an instance of suspicious or malicious activity, possibly involving LOLbins, that may be resistant to detection using conventional search rules.

Read More

A machine learning job has detected a suspicious Windows process. This process has been classified as malicious in two ways. It was predicted to be malicious by the ProblemChild supervised ML model, and it was found to be suspicious given that its user context is unusual and does not commonly manifest malicious activity,by an unsupervised ML model. Such a process may be an instance of suspicious or malicious activity, possibly involving LOLbins, that may be resistant to detection using conventional search rules.

Read More

A machine learning job has detected a rare process writing data to an external device. Malicious actors often use benign-looking processes to mask their data exfiltration activities. The discovery of such a process that has no legitimate reason to write data to external devices can indicate exfiltration.

Read More

An anomaly detection job has detected a remote file transfer on an unusual directory indicating a potential lateral movement activity on the host. Many Security solutions monitor well-known directories for suspicious activities, so attackers might use less common directories to bypass monitoring.

Read More

An anomaly detection job has detected a remote file transfer with a rare extension, which could indicate potential lateral movement activity on the host.

Read More

A machine learning job has detected an unusually high file size shared by a remote host indicating potential lateral movement activity. One of the primary goals of attackers after gaining access to a network is to locate and exfiltrate valuable information. Instead of multiple small transfers that can raise alarms, attackers might choose to bundle data into a single large file transfer.

Read More

A machine learning job has detected an RDP session started at an usual time or weekday. An RDP session at an unusual time could be followed by other suspicious activities, so catching this is a good first step in detecting a larger attack.

Read More