Palo Alto Networks, Inc.

Netskope, Inc.

Zhi Xu

Cong Zheng

Wenjun Hu

Xiao Meng Zhang

Zihang Xiao

Yihua Liao

Ari Azarafrooz

Najmeh Miramirkhani

Jiajie Wang

Huagang Xie

Xinran Wang

Tongbo Luo

Lin Xu

Shadi Rostami-Hesarsorkh

Seokkyung Chung

Farshad Rostamabadi

Elad Wexler

Lee Klarich

Asaf Weiss

William Redington Hewlett, Ii

Yangjie Chang

William Hewlett

Yangjie Cheng

A set of repackaging fingerprints generated independently of a particular original application is received. The set of repackaging fingerprints comprises a plurality of predetermined indicators of build-related structure that is independent of the particular original application's code structure. A mobile application is received. The received mobile application is analyzed for one or more indicators that the received mobile application is a repackaged version of the particular original application, using at least one repackaging fingerprint. In response to a result of the analysis, the received mobile application is categorized as a repackaged application.

Detecting repackaged applications based on file format fingerprints

Disclosed is phishing classifier that classifies a URL and content page accessed via the URL as phishing or not is disclosed, with URL feature hasher that parses and hashes the URL to produce feature hashes, and headless browser to access and internally render a content page at the URL, extract HTML tokens, and capture an image of the rendering. Also disclosed are an HTML encoder, trained on HTML tokens extracted from pages at URLs, encoded, then decoded to reproduce images captured from rendering, that produces an HTML encoding of the tokens extracted, and an image embedder, pretrained on images, that produces an image embedding of the image captured. Further, phishing classifier layers, trained on the feature hashes, the HTML encoding, and the image embedding, process the URL feature hashes, HTML encoding and image embeddings to produce a likelihood score that the URL and the page accessed presents a phishing risk.

Detecting phishing websites via a machine learning-based system using URL feature hashes, HTML encodings and embedded images of content pages

An indication of an application to be installed on a local device is received. A request is transmitted to a remote server for information associated with the application. In some cases, in response to the receipt of a report from the remote server, a set of rules restricting behaviors of the application is implemented at the local device. In some cases, in response to the receipt of a report from the remote server, the installation of the application on the local device is prevented.

Integrated application analysis and endpoint protection

Execution of an application in an application-level sandbox is disclosed. A request to launch an application is received by an operating system executing on a device. A determination is made that a stored copy of the application should be executed within an application-level sandbox. The stored copy of the application is executed in the application-level sandbox.

Application-level sandboxing on devices

A set of metadata associated with a plurality of samples is received. The samples are clustered. For members of a first cluster, a set of similarities shared among at least a portion of the members of the first cluster is determined. A cluster member is identified within the first cluster, and in response, additional analysis is caused to be performed on the outlier cluster member.

Automated malware family signature generation

Network traffic collectively associated with a set of communications made between a client device and a respective first and second resource during a usage of a program executed on the client device is monitored. At least of the monitored traffic associated with the second resource is encrypted. The encrypted traffic is categorized. A respective first and second policy to apply to the respective first and second communications is determined. The second policy is based at least in part on the categorization of the encrypted traffic.

Identifying security risks and enforcing policies on encrypted/encoded network communications

Evaluating samples is disclosed. A sample is received. A system component dependency graph is built for the sample. The system component dependency graph includes a plurality of nodes and at least one edge. A first node included in the plurality of nodes is one of: (1) a system component, (2) an indirect call component, or (3) a system event. The edge is an indirect call instruction. A verdict for the sample is determined based at least in part on the system component dependency graph.

Context-based analysis of applications

A sample is analyzed to determine a set of events that should be selected for performing by a dynamic analyzer executing the sample in an instrumented, emulated environment. The set of selected events is performed. In some cases, at least one emulator detection resistance action is performed. A maliciousness verdict is determined for the sample based at least in part on one or more responses taken by the sample in response to the set of selected events being performed by the dynamic analyzer.

Dynamic analysis techniques for applications

The automatic generation of malware family signatures is disclosed. A set of metadata associated with a plurality of samples is received. The samples are clustered. For members of a first cluster, a set of similarities shared among at least a portion of the members of the first cluster is determined. The similarities are evaluated for suitability as a malware family signature. Suitability is evaluated based on how well the similarities uniquely identify the members of the first cluster. In the event the similarities are determined to be suitable as a malware family signature, a signature is generated.

A sample is analyzed to determine a set of events that should be selected for performing by a dynamic analyzer executing the sample in an instrumented, emulated environment. In some cases, analyzing the sample includes extracting the sample's user interface layout into a tree hierarchy of user interface elements. The set of selected events is performed. In some cases, at least one emulator detection resistance action is performed. A maliciousness verdict is determined for the sample based at least in part on one or more responses taken by the sample in response to the set of selected events being performed by the dynamic analyzer.

A virtualized storage for use in performing dynamic analysis on a sample is configured, at least in part by copying the sample to the virtualized storage. A virtual machine emulator is launched using a snapshot of a virtualized platform. The virtualized platform is previously configured to use the virtualized storage, and the snapshot is configured to use a placeholder file to occupy space for later use when installing the sample. A location of the copied sample in an image corresponding to the virtualized storage is determined. The copied sample is installed and dynamic analysis is performed on the sample.

Disclosed is classifying a URL and a page accessed via the URL as phishing or not. URL embedder extracts characters in a predetermined set from the URL to produce a character string trained using ground truth classification of the URL, producing a URL embedding. HTML parser accesses content at the URL and extracts HTML tokens from the page. Further, HTML encoder, trained on HTML tokens extracted from pages at example URLs, each example URL accompanied by a ground truth image captured from the page accessed via the example URL, produces an HTML encoding of the extracted tokens. Also, phishing classifier layers, trained on the URL embedding and the HTML encoding of example URLs, processes a concatenated input of the URL embedding and the HTML encoding to produce a score of a phishing risk.

Machine learning-based system for detecting phishing websites using the URLS, word encodings and images of content pages

Disclosed is classifying a URL and a content page accessed via the URL as phishing or not. URL embedder extracts characters in a predetermined set from the URL to produce a character string trained using ground truth classification of the URL, producing a URL embedding. HTML parser accesses content at the URL and extracts HTML tokens from the content page. Further, HTML encoder, trained on HTML tokens extracted from content pages at example URLs, each example URL accompanied by a ground truth image captured from the content page accessed via the example URL, produces an HTML encoding of the tokens extracted from the page. Also, phishing classifier layers, trained on the URL embedding and the HTML encoding of example URLs, processes a concatenated input of the URL embedding and the HTML encoding to produce a likelihood score that the URL and content accessed via the URL presents a phishing risk.

Growing community of inventors

Cupertino, CA, United States of America

Zhi Xu