Graph-based detection of lateral movement in computer networks
Systems and methods are disclosed to implement a cyberattack detection system that monitors a computer network for lateral movement. In embodiments, the system uses network data from a computer network to build a baseline of connection behaviors for the network. Connection graphs are generated from new network data that indicate groups of nodes that made connections with one another during a last time interval. The graphs are analyzed for connection behavior anomalies and ranked to determine a subset of graphs with suspected lateral movement. Graphs with suspected lateral movement may be further analyzed to determine a set of possible attack paths in the lateral movements. The suspected attack paths are reported to network administrators via a notification interface. Advantageously, the disclosed system is able to detect potential lateral movements in localized portions of a network by monitoring for connection behavior anomalies in network data gathered from the network.
1 . A system, comprising: one or more hardware processors with associated memory that implement a cyberattack detection system, configured to:
collect network data from a computer network, wherein the network data indicates connections made or attempted among nodes of the computer network during a time interval and respective times that the connections were made or attempted;
generate a plurality of connection graphs from the network data, wherein each of the connection graphs indicates a subset of the nodes of the computer network that made or attempted connections with one another during the time interval and one or more anomalies in connection behavior of the subset of nodes during the time interval;
compute a graph metric for each of the connection graphs based on a highest degree of nodes in individual ones of the connection graphs and determine, based on the graph metric, a set of highest-ranking connection graphs to evaluate for a suspected lateral movement of a cyberattack during the time interval;
for each connection graph in the set of highest-ranking connection graphs:
determine a connection path from the connection graph, wherein the connection path indicates a series of nodes linked by a sequence of time-ordered connections and one or more anomalies in connection behavior of the series of nodes during the time interval; and
compute a path metric of the connection path and determine, based on the path metric, that the connection path is a suspected attack path of the suspected lateral movement; and
provide an output that indicates the suspected attack paths during the time interval.
2 . The system of claim 1 , wherein the cyberattack detection system is configured to:
determine the set of highest-ranking connection graphs based on a specified number of connection graphs for the set.
3 . The system of claim 1 , wherein the cyberattack detection system is configured to:
generate a plurality of connection paths of time-ordered connections from the connection graph, including the connection path;
compute respective path metrics of the connection paths;
rank the connection paths according to their respective path metrics; and
wherein the determination that the connection path is a suspected attack path is made based on a rank of the connection path among the plurality of connection paths.
4 . The system of claim 1 , wherein the cyberattack detection system is configured to:
determine, from the network data, the one or more connection behavior anomalies indicated in the connection graph, including one or more of:
a node making a connection using a new protocol during the time interval,
a node making a connection to a new destination node during the time interval, and
a node making more connection during the time interval compared to one or more previous time intervals; and
wherein the graph metric is computed based on the one or more connection behavior anomalies.
5 . The system of claim 1 , wherein the cyberattack detection system is configured to:
determine that a particular node in the computer network is a scanner node configured to attempt connections to a plurality of other nodes; and
exclude the scanner node from the generation of the connection graph.
6 . The system of claim 1 , wherein the cyberattack detection system is configured to:
determine that a particular connection in the network data is a whitelisted connection; and
exclude the whitelisted connection from the generation of the connection graph.
7 . The system of claim 1 , wherein the cyberattack detection system is implemented with resources of a platform-as-a-service (PaaS) provider, and comprises:
a network data collection service configured to collect network data from a plurality of client networks; and
an attack monitoring service configured to programmatically generate alerts indicating suspected lateral movement in the client networks.
8 . A method, comprising: performing, by one or more computers that implement a cyberattack detection system:
collecting network data from a computer network indicating connections made or attempted among nodes of the computer network during a time interval and respective times that the connections were made or attempted;
generating a plurality of connection graphs from the network data, wherein each of the connection graphs indicates a subset of the nodes of the computer network that made or attempted connections with one another during the time interval and one or more anomalies in connection behavior of the subset of nodes during the time interval;
computing a graph metric for each of the connection graphs based on a highest degree of nodes in individual ones of the connection graphs and determining, based on the graph metric, a set of highest-ranking connection graphs to evaluate for a suspected lateral movement of a cyberattack during the time interval;
determine, based on the graph metric, a set of highest-ranking connection graphs to evaluate for a suspected lateral movement of a cyberattack during the time interval;
for each connection graph in the set of highest-ranking connection graphs:
determining a connection path from the connection graph, wherein the connection path indicates a series of nodes linked by a sequence of time-ordered connections and one or more anomalies in connection behavior of the series of nodes during the time interval; and
computing a path metric of the connection path and determining, based on the path metric, that the connection path is a suspected attack path of the suspected lateral movement; and
generating an output indicating the suspected attack paths during the time interval.
9 . The method of claim 8 , wherein:
the network data includes connections of different communication protocols, including two or more of:
Secure Shell (SSH) protocol,
Windows Remote Management (WinRM) protocol,
Remote Desktop Protocol (RDP), and
Server Message Block (SMB) protocol; and
the connection graph indicates the different communication protocols.
10 . The method of claim 8 , wherein:
the graph metric is computed based on two or more of:
an order of the connection graph,
a size of the connection graph,
a diameter of the connection graph,
a number of distinct communication protocols included in the connection graph,
a presence or count of new nodes in the connection graph, and a presence or count of new edges in the connection graph.
11 . The method of claim 8 , wherein:
the path metric is computed based on two or more of:
a number of links in the connection path,
a number of distinct communication protocols included in the connection path,
a presence or count of new nodes in the connection path, and
a presence or count of new links in the connection path.
12 . The method of claim 8 , further comprising performing, by the cyberattack detection system:
repeatedly generating a plurality of connection graphs from the network data periodically based on a specified schedule.
13 . The method of claim 8 , the generating of the connection graph is triggered by a detection that one or more nodes in the network data exceeded an activity level threshold during the time interval.
14 . The method of claim 8 , further comprising performing, by the cyberattack detection system:
computing, at successive time intervals, a network behavior tracking metric based on a plurality path metrics of suspected attack paths determined during each time interval;
determining that a difference between the network behavior tracking metric in one time interval and a historical average of the network behavior tracking metric triggers an alert criterion; and
generating an alert responsive to the triggering of the alert criterion.
15 . The method of claim 8 , wherein:
the output is provided via a graphical user interface of the cyberattack detection system; and
the output includes the connection graph and the connection path.
16 . The method of claim 8 , wherein:
the output is provided via a notification interface of the cyberattack detection system that sends one or more notifications of the suspected lateral movement to one or more registered user accounts.
17 . The method of claim 8 , further comprising performing, by the cyberattack detection system:
receiving, via a configuration interface, configuration information specifying one or more graph properties to use to compute the graph metric; and
receiving, via the configuration interface, additional configuration information specifying one or more path properties to use to compute the path metric.
18 . One or more non-transitory computer-accessible storage media storing program instructions that when executed on or across one or more processors implement at least a portion of a cyberattack detection system and cause the cyberattack detection system to:
analyze network data collected from a computer network, wherein the network data indicates connections made or attempted among nodes of the computer network during a time interval and respective times that the connections were made or attempted;
generate a plurality of connection graphs from the network data, wherein each of the connection graphs indicates a subset of the nodes of the computer network that made or attempted connections with one another during the time interval and one or more anomalies in connection behavior of the subset of nodes during the time interval;
compute a graph metric for each of the connection graphs based on a highest degree of nodes in individual ones of the connection graphs and determine, based on the graph metric, a set of highest-ranking connection graphs to evaluate for a suspected lateral movement of a cyberattack during the time interval;
for each connection graph in the set of highest-ranking connection graphs:
determine a connection path from the connection graph, wherein the connection path indicates a series of nodes linked by a sequence of time-ordered connections and one or more anomalies in connection behavior of the series of nodes during the time interval; and
compute a path metric of the connection path and determine, based on the path metric, that the connection path is a suspected attack path of the suspected lateral movement; and
generate output that indicates the suspected attack paths during the time interval.
19 . The one or more non-transitory computer- accessible storage media of claim 18 , wherein the program instructions when executed on or across the one or more processors cause the cyberattack detection system to: determine the set of highest-ranking connection graphs based on a specified number of connection graphs for the set.
20 . The one or more non-transitory computer-accessible storage media of claim 18 , wherein the program instructions when executed on or across the one or more processors cause the cyberattack detection system to:
compute, at successive time intervals, a network behavior tracking metric based on a plurality path metrics of suspected attack paths determined during each time interval;
determine that a difference between the network behavior tracking metric in one time interval and a historical average of the network behavior tracking metric triggers an alert criterion; and
generate an alert responsive to the triggering of the alert criterion.