Data Mining Practical Machine Learning Tools and Techniques

Dynamic Taint Analysis for Automatic Detection, Analysis, and Signature Generation of Exploits on Commodity Software, Network and Distributed System Security Symposium Conference Proceedings : 2005

We present DNADroid, a tool that detects Android application copying, or “cloning”, by robustly computing the similarity between two applications. DNADroid achieves this by comparing program dependency graphs between methods in candidate applications. Using DNADroid, we found at least 141 applications that have been the victims of cloning, some as many as seven times. DNADroid has a very low false positive rate — we manually confirmed that all the applications detected are indeed clones by either visual or behavioral similarity. We present several case studies that give insight into why applications are cloned, including localization and redirecting ad revenue. We describe a case of malware being added to an application and show how DNADroid was able to detect two variants of the same malware. Lastly, we offer examples of an open source cracking tool being used in the wild.

/pdf/attack-of-the-clones-detecting-cloned-applications-on-2wglc561qa.pdf

Attack of the Clones: Detecting Cloned Applications on Android Markets

Besides traditional problems such as potential bugs, (smartphone) application clones on Android markets bring new threats. That is, attackers clone the code from legitimate Android applications, assemble it with malicious code or advertisements, and publish these ``purpose-added" app clones on the same or other markets for benefits. Three inherent and unique characteristics make app clones difficult to detect by existing techniques: a billion opcode problem caused by cross-market publishing, gap between code clones and app clones, and prevalent Type 2 and Type 3 clones. Existing techniques achieve either accuracy or scalability, but not both. To achieve both goals, we use a geometry characteristic, called centroid, of dependency graphs to measure the similarity between methods (code fragments) in two apps. Then we synthesize the method-level similarities and draw a Y/N conclusion on app (core functionality) cloning. The observed ``centroid effect" and the inherent ``monotonicity" property enable our approach to achieve both high accuracy and scalability. We implemented the app clone detection system and evaluated it on five whole Android markets (including 150,145 apps, 203 million methods and 26 billion opcodes). It takes less than one hour to perform cross-market app clone detection on the five markets after generating centroids only once.

/pdf/achieving-accuracy-and-scalability-simultaneously-in-9lujnp45g3.pdf

Achieving accuracy and scalability simultaneously in detecting application clones on Android markets

An electronic message is analyzed for malware contained in the message. Text of an electronic message may be analyzed to detect and process malware content in the electronic message itself. The present technology may analyze an electronic message and attachments to electronic messages to detect a uniform resource location (URL), identify whether the URL is suspicious, and analyze all suspicious URLs to determine if they are malware. The analysis may include re-playing the suspicious URL in a virtual environment which simulates the intended computing device to receive the electronic message. If the re-played URL is determined to be malicious, the malicious URL is added to a black list which is updated throughout the computer system.

Electronic message analysis for malware detection

A proactive worm containment (PWC) solution for enterprises uses a sustained faster-than-normal outgoing connection rate to determine if a host is infected. Two novel white detection techniques are used to reduce false positives, including a vulnerability time window lemma to avoid false initial containment, and a relaxation analysis to uncontain (or unblock) those mistakenly contained (or blocked) hosts, if there are any. The system integrates seamlessly with existing signature-based or filter-based worm scan filtering solutions. Nevertheless, the invention is signature free and does not rely on worm signatures. Nor is it protocol specific, as the approach performs containment consistently over a large range of worm scan rates. It is not sensitive to worm scan rate and, being a network-level approach deployed on a host, the system requires no changes to the host's OS, applications, or hardware.

Proactive worm containment (pwc) for enterprise networks

Along with the burst of open source projects, software theft (or plagiarism) has become a very serious threat to the healthiness of software industry. Software birthmark, which represents the unique characteristics of a program, can be used for software theft detection. We propose a system call dependence graph based software birthmark called SCDG birthmark, and examine how well it reflects unique behavioral characteristics of a program. To our knowledge, our detection system based on SCDG birthmark is the first one that is capable of detecting software component theft where only partial code is stolen. We demonstrate the strength of our birthmark against various evasion techniques, including those based on different compilers and different compiler optimization levels as well as two state-of-the-art obfuscation tools. Unlike the existing work that were evaluated through small or toy software, we also evaluate our birthmark on a set of large software. Our results show that SCDG birthmark is very practical and effective in detecting software theft that even adopts advanced evasion techniques.

/pdf/behavior-based-software-theft-detection-jd7ik5xi4y.pdf

Behavior based software theft detection

Along with the burst of open source projects, software theft (or plagiarism) has become a very serious threat to the healthiness of software industry. Software birthmark, which represents the unique characteristic of a program, can be used for software theft detection. We propose two system call based software birthmarks: SCSSB (System Call Short Sequence Birthmark) and IDSCSB (Input Dependant System Call Subsequence Birthmark), and examine how well they reflect unique behavioral characteristics of a program. To our knowledge, our detection system based on SCSSB and IDSCSB is the first one that is capable of software component theft detection where only partial code is stolen. We demonstrate the strength of our birthmarks against various evasion techniques, including those based on different compilers and different compiler optimization levels as well as those based on very powerful obfuscation techniques supported by SandMark. Unlike the existing work that were evaluated through small or toy software, we also evaluate our birthmarks on a set of large software (web browsers). Our results show that system call based birthmarks are very practical and effective in detecting software theft that even adopts advanced evasion techniques.

Detecting Software Theft via System Call Based Birthmarks

Identifying similar or identical code fragments becomes much more challenging in code theft cases where plagiarizers can use various automated code transformation techniques to hide stolen code from being detected. Previous works in this field are largely limited in that (1) most of them cannot handle advanced obfuscation techniques; (2) the methods based on source code analysis are less practical since the source code of suspicious programs is typically not available until strong evidences are collected; and (3) those depending on the features of specific operating systems or programming languages have limited applicability. Based on an observation that some critical runtime values are hard to be replaced or eliminated by semantics-preserving transformation techniques, we introduce a novel approach to dynamic characterization of executable programs. Leveraging such invariant values, our technique is resilient to various control and data obfuscation techniques. We show how the values can be extracted and refined to expose the critical values and how we can apply this runtime property to help solve problems in software plagiarism detection. We have implemented a prototype with a dynamic taint analyzer atop a generic processor emulator. Our experimental results show that the value-based method successfully discriminates 34 plagiarisms obfuscated by SandMark, plagiarisms heavily obfuscated by KlassMaster, programs obfuscated by Thicket, and executables obfuscated by Loco/Diablo.

/pdf/value-based-program-characterization-and-its-application-to-5aixawi2sm.pdf

Value-based program characterization and its application to software plagiarism detection

In this work, we address the problem of algorithm plagiarism, which occurs when a plagiarist, violating intellectual property rights, steals others' algorithms and covertly implements them. In contrast to software plagiarism, which has been extensively studied, limited attention has been paid to algorithm plagiarism. In this paper, we propose two dynamic value-based approaches, namely N-version and annotation, for algorithm plagiarism detection. Our approaches are motivated by the observation that there exist some critical runtime values which are irreplaceable and uneliminatable for all implementations of the same algorithm. The N-version approach extracts such values by filtering out non-core values. The annotation approach leverages auxiliary information to flag important variables which contain core values. We also propose a value dependence graph based similarity metric in addition to the longest common subsequence based one, in order to address the potential value reordering attack. We have implemented a prototype and evaluated the proposed schemes on various algorithms. The results show that our approaches to algorithm plagiarism detection are practical, effective and resilient to many automatic obfuscation techniques.

/pdf/a-first-step-towards-algorithm-plagiarism-detection-2teoe6o6um.pdf

Yoon-Chan Jhi

Papers

Proactive worm containment (pwc) for enterprise networks

Behavior based software theft detection

Detecting Software Theft via System Call Based Birthmarks

Value-based program characterization and its application to software plagiarism detection

A first step towards algorithm plagiarism detection