Soteris Demetriou, Whitney Merrill, Wei Yang, Aston Zhang and Carl A

Slides:



Advertisements
Similar presentations
Information System Audit : © South-Asian Management Technologies Foundation Chapter 4: Information System Audit Requirements.
Advertisements

Android architecture overview
Spyware and Adware Rick Carback 9/18/2005
 Firewalls and Application Level Gateways (ALGs)  Usually configured to protect from at least two types of attack ▪ Control sites which local users.
Methods For The Prevention, Detection And Removal Of Software Security Vulnerabilities Jay-Evan J. Tevis Department of Computer Science and Software Engineering.
Android Security Enforcement and Refinement. Android Applications --- Example Example of location-sensitive social networking application for mobile phones.
3-1 Chapter Three. 3-2 Secondary Data vs. Primary Data Secondary Data: Data that have been gathered previously. Primary Data: New data gathered to help.
Instant Messaging Security Flaws By: Shadow404 Southern Poly University.
Presentation By Deepak Katta
Understanding Android Security Yinshu Wu William Enck, Machigar Ongtang, and PatrickMcDaniel Pennsylvania State University.
Introduction Our Topic: Mobile Security Why is mobile security important?
A METHODOLOGY FOR EMPIRICAL ANALYSIS OF PERMISSION-BASED SECURITY MODELS AND ITS APPLICATION TO ANDROID.
Drupal Security Securing your Configuration Justin C. Klein Keane University of Pennsylvania School of Arts and Sciences Information Security and Unix.
Presented by: Kushal Mehta University of Central Florida Michael Spreitzenbarth, Felix Freiling Friedrich-Alexander- University Erlangen, Germany michael.spreitzenbart,
1 All Your iFRAMEs Point to Us Mike Burry. 2 Drive-by downloads Malicious code (typically Javascript) Downloaded without user interaction (automatic),
What is FORENSICS? Why do we need Network Forensics?
SUPOR : Precise and Scalable Sensitive User Input Detection for Android Apps Jianjun Huang, Zhichun Li, Xusheng Xiao, Zhenyu Wu, Kangjie Lu, Xiangyu Zhang,
App Rights or wrongs ? A look at smartphone apps or: why RTFM* is not just important for geeks and “computer types” * = Read The F+*#ing (or “Fine”) Manual.
Lecture slides prepared for “Computer Security: Principles and Practice”, 3/e, by William Stallings and Lawrie Brown, Chapter 5 “Database and Cloud Security”.
CS378 - Mobile Computing Intents.
Android for Java Developers Denver Java Users Group Jan 11, Mike
Smart Machines, Smart Privacy: Rules of the Road and Challenges Ahead The views expressed are those of the speaker and not necessarily those of the FTC.
CS378 - Mobile Computing Intents. Allow us to use applications and components that are part of Android System – start activities – start services – deliver.
CC3020N Fundamentals of Security Management CC3020N Fundamentals of Security Management Lecture 2 Risk Identification and Risk Assessment.
Lecture 31 Risk Management. Introduction Information security departments are created primarily to manage IT risk Managing risk is one of the key responsibilities.
Evaluation of Spam Detection and Prevention Frameworks for and Image Spam - A State of Art Pedram Hayati, Vidyasagar Potdar Digital Ecosystems and.
ADV. NETWORK SECURITY CODY WATSON What’s in Your Dongle and Bank Account? Mandatory and Discretionary Protections of External Resources.
BioSumm A novel summarizer oriented to biological information Elena Baralis, Alessandro Fiori, Lorenzo Montrucchio Politecnico di Torino Introduction text.
Leave Me Alone: App- level Protection Against Runtime Information Gathering on Android NAN ZHANG, KAN YUAN, MUHAMMAD NAVEED†, XIAOYONG ZHOU AND XIAOFENG.
November 19, 2008 CSC 682 Use of Virtualization to Thwart Malware Written by: Ryan Lehan Presented by: Ryan Lehan Directed By: Ryan Lehan Produced By:
1 Lab 12: Spyware A Window’s User’s Worst Nightmare.
Wireless and Mobile Security
IS493 INFORMATION SECURITY TUTORIAL # 1 (S ) ASHRAF YOUSSEF.
Identifying “Best Bet” Web Search Results by Mining Past User Behavior Author: Eugene Agichtein, Zijian Zheng (Microsoft Research) Source: KDD2006 Reporter:
Power Guru: Implementing Smart Power Management on the Android Platform Written by Raef Mchaymech.
©2015 Check Point Software Technologies Ltd. 1 [Restricted] ONLY for designated groups and individuals CHECK POINT MOBILE THREAT PREVENTION.
Android and IOS Permissions Why are they here and what do they want from me?
What’s in Your Dongle and Bank Account? Mandatory and Discretionary Protection of Android External Resources Literature by S. Demetriou et al. Presented.
Threat Modeling: Employing the 5 Ws Security Series, December 13, 2013 Jeff Minelli Penn State ITS
Lecturer: Eng. Mohamed Adam Isak PH.D Researcher in CS M.Sc. and B.Sc. of Information Technology Engineering, Lecturer in University of Somalia and Mogadishu.
Computer Security Keeping you and your computer safe in the digital world.
Using Honeypots to Improve Network Security Dr. Saleh Ibrahim Almotairi Research and Development Centre National Information Centre - Ministry of Interior.
Preparing Your Apps for Publication Test your app thoroughly on a variety of devices. The app might work perfectly using the emulator on your.
Profiling: What is it? Notes and reflections on profiling and how it could be used in process mining.
ANDROID ACCESS CONTROL Presented by: Justin Williams Masters of Computer Science Candidate.
What mobile ads know about mobile users
Module 51 (Mobile Device Fundamentals - Android)
SIEM Rotem Mesika System security engineering
Database and Cloud Security
Advanced Endpoint Security Data Connectors-Charlotte January 2016
Botnets A collection of compromised machines
The Price of Free Privacy Leakage in Personalized Mobile In-App Ads
What Mobile Ads know about mobile users
Free for All! Assessing User Data Exposure to Advertising libraries on Android John Ramirez.
Free for All! Assessing User Data Exposure to Advertising Libraries on Android Campbell Foskin.
Presentation by Jun Hao Xu
Understanding Android Security
Designing Cross-Language Information Retrieval System using various Techniques of Query Expansion and Indexing for Improved Performance  Hello everyone,
Are these ads safe? Detecting hidden attacks through the mobile app-web interface Vaibhav Rastogi, Rui Shao, Yan Chen, Xiang Pan, Shihong Zou, and Ryan.
Presented by Xiaohui (Amy) Lin
Web Mining Ref:
TriggerScope Towards Detecting Logic Bombs in Android Applications
What Mobile Ads Know About Mobile Users
Botnets A collection of compromised machines
Xutong Chen and Yan Chen
Network Profiler: Towards Automatic Fingerprinting of Android Apps
Mobile App Advertisements
Understanding Android Security
Mobile Security What is mobile secuirty & Identifying smartphone security holes& Sayed Hashimi Proposal Project.
When Machine Learning Meets Security – Secure ML or Use ML to Secure sth.? ECE 693.
Presentation transcript:

Free for All! Assessing User Data Exposure to Advertising Libraries on Android Soteris Demetriou, Whitney Merrill, Wei Yang, Aston Zhang and Carl A. Gunter Christina Bell cbel296 - 8969895

introduction Advertisers aim to generate ad conversions for their ad impressions Ad networks help by matching ads to users Assess potential risks – all possible behaviours (not only current behaviour) 4 major attack channels Unprotected APIs Protected APIs Access to host app files Observing user input Developed Pluto framework Analyse app and help developer assess potential risks

Background: Mobile Advertising Many developers monetise their apps through ads. Data brokers: incorporate ad libraries through applications collect targeted data - user attributes and interests sell the user profiles to the advertising companies Data brokers collaborate with advertisers to create more suitable ads. Accurate data → targeting correct users → more clicks → $$$

Background: android Each app has unique UID and PID – which extends to the ad library Host apps share their privileges and resources with its ad libraries Linux DAC security system allows the ad library to access files generated by host app. Ad libraries have already been collating user info without user knowledge.

Background: NLP Data miners use NLP to determine if words are data points and to determine which part of speech each word is. Targeted data can be vague so NLP is used to determine the semantic meaning. Word net (English semantic dictionary) Similarity metric used to determine if words are associated.

Threat model Risk: potential compromise of an asset through the exploit of a vulnerability done by a threat. Asset – targeted data Vulnerability – What allows ad libraries to gain sensitive information. Threat – opportunistic ad library Attack channels are divided into categories In-app: dependent on the ad library host app Out-app: independent of the host app

In-app channels Ad libraries can leverage their position within their host app to access exposed data 1. Parse local files generated by host app at runtime 2. Inherit the permissions granted to its host app 3. Peek on host app user input Manual inspection of real world free apps Several data points were found to be exposed. “I’m Pregnant” (1-5 million downloads) exposed weight, height, current pregnancy month and day “TalkLife” (10-50 thousand downloads) exposed email address, birth date, first name, password in plain text

In-app channels cont. Level One Inspection (L1-I): Attack technique that examines local files and protected APIs. Manual inspection of 262 applications Level Two Inspection (L2-I): Attack technique that utilises L1-1 as well as eavesdropping on user input. Manual inspection of 35 applications.

Out-app channels Access targeted data independently of their host application Public APIs that return app bundles. getInstalledPackages() getInstalledApplications() 12.54% of apps examined incorporate ad libraries that called either of these methods. Issue as user is not informed. Children apps don’t get permission of parent. E.g. Radio Disney

Pluto framework Modular framework for estimating in-app and out-app targeted data exposure of an app. In-app Pluto focuses on analysing local files generated by the host app. Layout file Resource file Manifest file Runtime generated files Out-app Pluto uses app bundles. Predicts which apps will be installed together Machine learning to draw inferences

In-app pluto Dynamic analysis module (DAM) File miners Runs host app in emulator Decompiles app and extracts file File miners User attributes and interests as a matching goal Goal reached when data point present in file Content disambiguation layer uses NLP similarity metric to determine whether to accept match - droidLESK

Out app pluto Out-app Pluto aims to estimate what is the potential data exposure to an ad library that uses getInstalledPackages() and getInstalledApplications() Explore what data points can be exposed from this list Co-installation patterns (CIP) Frequent pattern mining (FPM) to find application co-installation patterns. Confidence Facebook => skype, viber with 70% confidence. Supervised learning Infer user attributes from the CIP estimated app bundles Classifiers used to class patterns found to relevant data points.

Criticism/recommendations Pluto NLP similarity metric takes advantage of software best practices. When extracting words camel case and snake case are assumed to be used. E.g. userProfile and user_profile detected but not userprof, uProf or up. This means more research into other naming conventions is needed. Research was limited to the four attack channels discussed. This could be expanded into other attack areas, such as camera, gyroscope, accelerometer or audio. Size of the study was limited to the 2535 apps that were analysed. This is small compared to the 2.8 million apps on the Google Play store. [1] [1]https://www.statista.com/statistics/266210/number-of-available-applications-in-the-google-play-store/

Criticism/recommendations The ‘risk’ of an app that is reported by Pluto is generalised. Each user is different. User has no input to what data points they consider as “risky” If an app crashed it was simply removed from the study No investigation Obfuscated applications or hidden ad libraries could pose a problem for Pluto Pluto had a large number of false positives in evaluation Tried to combat with droidLESK

Questions?