Cisco Technology, Inc.

Tomas Pevny

David A McGrew

Jan Kohout

Martin Kopp

Martin Grill

Martin Rehak

Petr Somol

Jan Jusko

Blake Harrell Anderson

Daniel George Wing

Andrew Zawadowskiy

Saravanan Radhakrishnan

Donovan O'Hara

Karel Bartos

Ivan Nikolaev

Martin Vejman

David Mcgrew

Jan Stiborek

Gustav Sourek

Filip Zelezny

Jan Franco̊

In one embodiment, a device in a network detects an encrypted traffic flow associated with a client in the network. The device captures contextual traffic data regarding the encrypted traffic flow from one or more unencrypted packets associated with the client. The device performs a classification of the encrypted traffic flow by using the contextual traffic data as input to a machine learning-based classifier. The device generates an alert based on the classification of the encrypted traffic flow.

Analyzing encrypted traffic behavior using contextual traffic data

A method comprises receiving, at a network infrastructure device, a flow of packets, determining, using the network infrastructure device and for a first subset of the packets, that the first subset corresponds to a first datagram and determining a first length of the first datagram, determining, using the network infrastructure device and for a second subset of the packets, that the second subset corresponds to a second datagram that was received after the first datagram, and determining a second length of the second datagram, determining, using the network infrastructure device, a duration value between a first arrival time of the first datagram and a second arrival time of the second datagram, sending, to a collector device that is separate from the network infrastructure device, the first length, the second length, and the duration value for analysis.

Detection of malware and malicious applications

In one embodiment, a device disassembles an executable file into assembly instructions. The device maps each of the assembly instructions to a fixed length instruction vector using one-hot encoding and an instruction vocabulary and forms vector representations of blocks of a control flow graph for corresponding functions of the executable file by embedding and aggregating bags of the instruction vectors. The device generates, based on the vector representations of the blocks of the control flow graph, a call graph model of the functions in the executable file. The device forms a vector representation of the executable file based in part on the call graph model. The device determines, based on the vector representation of the executable file, whether the executable file is malware.

Detection of malicious executable files using hierarchical models

In one embodiment, a device analyzes network traffic data using a clustering process, to identify a cluster of addresses associated with the network traffic data for which the associated network traffic has similar behavioral characteristics. The device calculates a set of rankings for the cluster by comparing the cluster to different sets of malicious addresses. The device aggregates the set of rankings into a final ranking by setting the rankings in the set as current rankings and iteratively calculating an average of any subset of the current rankings that comprises correlated rankings. The calculated average replaces the rankings in the subset as a current ranking. When none of the current rankings are correlated, the device performs an aggregation across all of the current rankings to form the final ranking. The device provides data indicative of the cluster for review by a supervisor, based on the final ranking.

Active prioritization of investigation targets in network security

In one embodiment, a device obtains simulation environment data regarding traffic generated within a simulation environment in which malware is executed. The device trains a malware detector using the simulation environment data. The device obtains deployment environment characteristics of a network to which the malware detector is to be deployed. The device configures the malware detector to ignore data in the simulation environment data that is associated with one or more environment characteristics that are not present in the deployment environment characteristics.

Leveraging endpoint and network environment inferences for malware traffic classification

In one embodiment, a device in a network identifies an set of services of a domain accessed by a plurality of users in the network. The device generates a service usage model for the domain based on the set of services accessed by the plurality of users. The service usage model models usage of the services of the domain by the plurality of users. The device trains a machine learning-based classifier to analyze traffic in the network using a set of training feature vectors. A particular training feature vector includes data indicative of service usage by one of the users for the domain and the modeled usage of the services of the domain by the plurality of users. The device causes classification of traffic in the network associated with a particular user by the trained machine learning-based classifier.

Service usage model for traffic analysis

In one embodiment, a device in a network maintains a plurality of machine learning-based detectors for an intrusion detection system. Each detector is associated with a different portion of a feature space of traffic characteristics assessed by the intrusion detection system. The device provides data regarding the plurality of detectors to a user interface. The device receives an adjustment instruction from the user interface based on the data provided to the user interface regarding the plurality of detectors. The device adjusts the portions of the feature space associated with the plurality of detectors based on the adjustment instruction received from the user interface.

Private-learned IDS

Detecting illegitimate typosquatting with Internet Protocol (IP) information includes, at a computing device having connectivity to a network, obtaining a list of domains and filtering the list to generate a list of monitored domain strings. IP information is passively determined for domains associated with each of the monitored domain strings. A domain requested in network traffic for the network is identified as a candidate typosquatting domain and the candidate typosquatting domain is determined to be an illegitimate typosquatting domain based at least on the IP information. An action is initiated related to the illegitimate typosquatting domain.

Illegitimate typosquatting detection with internet protocol information

Presented herein are techniques for classifying devices as being infected with malware based on learned indicators of compromise. A method includes receiving at a security analysis device, traffic flows from a plurality of entities destined for a plurality of users, aggregating the traffic flows into discrete bags of traffic, wherein the bags of traffic comprise a plurality of flows of traffic for a given user over a predetermined period of time, extracting features from the bags of traffic and aggregating the features into per-flow feature vectors, aggregating the per-flow feature vectors into per-destination domain aggregated vectors, combining the per-destination-domain aggregated vectors into a per-user aggregated vector, and classifying a computing device used by a given user as infected with malware when indicators of compromise detected in the bags of traffic indicate that the per-user aggregated vector for the given user includes suspicious features among the extracted features.

Learning indicators of compromise with hierarchical models

In an embodiment, the method comprises receiving an identification of an anomaly associated with a false positive identification of a security threat by the intrusion detection system, wherein a first set of feature data identifies features of the anomaly; creating a plurality of training sets each comprising identifications of a plurality of samples of network communications; for the anomaly and each training set of the plurality of training sets, training a decision tree that is stored in digital memory of the security analysis computer; based at least in part on the plurality of trained decision trees, extracting a set of features that distinguish the anomaly from the plurality of samples; generating one or more rules associated with the anomaly from the extracted set of features and causing programming the security analysis computer with the one or more rules.

Explaining network anomalies using decision trees

Rapidly detecting network threats with targeted detectors includes, at a computing device having connectivity to a network, determining features of background network traffic. Features are also extracted from a particular type of network threat. A characteristic of the particular type of network threat that best differentiates the features of the particular type of network threat from the features of the background network traffic is determined. A targeted detector for the particular type of network threat is created based on the characteristic and an action is applied to particular incoming network traffic identified by the targeted detector as being associated with the particular type of network threat.

Growing community of inventors

Prague, Czechia

Tomas Pevny