Recent research efforts on adversarial ML have investigated problem-space attacks, focusing on the generation of real evasive objects in domains where, unlike images, there is no clear inverse mapping to the feature space (e.g., software). However, the design, comparison, and real-world implications of problem-space attacks remain underexplored.This paper makes two major contributions. First, we propose a novel formalization for adversarial ML evasion attacks in the problem-space, which includes the definition of a comprehensive set of constraints on available transformations, preserved semantics, robustness to preprocessing, and plausibility. We shed light on the relationship between feature space and problem space, and we introduce the concept of side-effect features as the byproduct of the inverse feature-mapping problem. This enables us to define and prove necessary and sufficient conditions for the existence of problem-space attacks. We further demonstrate the expressive power of our formalization by using it to describe several attacks from related literature across different domains.Second, building on our formalization, we propose a novel problem-space attack on Android malware that overcomes past limitations. Experiments on a dataset with 170K Android apps from 2017 and 2018 show the practical feasibility of evading a state-of-the-art malware classifier along with its hardened version. Our results demonstrate that "adversarial-malware as a service" is a realistic threat, as we automatically generate thousands of realistic and inconspicuous adversarial applications at scale, where on average it takes only a few minutes to generate an adversarial app. Yet, out of the 1600+ papers on adversarial ML published in the past six years, roughly 40 focus on malware [15]-and many remain only in the feature space.Our formalization of problem-space attacks paves the way to more principled research in this domain. We responsibly release the code and dataset of our novel attack to other researchers, to encourage future work on defenses in the problem space.

With the integration of mobile devices into daily life, smartphones are privy to increasing amounts of sensitive information. Sophisticated mobile malware, particularly Android malware, acquire or utilize such data without user consent. It is therefore essential to devise effective techniques to analyze and detect these threats. This article presents a comprehensive survey on leading Android malware analysis and detection techniques, and their effectiveness against evolving malware. This article categorizes systems by methodology and date to evaluate progression and weaknesses. This article also discusses evaluations of industry solutions, malware statistics, and malware evasion techniques and concludes by supporting future research paths.

Mobile devices and their application marketplaces drive the entire economy of the today's mobile landscape. Android platforms alone have produced staggering revenues, exceeding five billion USD, which has attracted cybercriminals and increased malware in Android markets at an alarming rate. To better understand this slew of threats, we present CopperDroid, an automatic VMI-based dynamic analysis system to reconstruct the behaviors of Android malware. The novelty of CopperDroid lies in its agnostic approach to identify interesting OS-and high-level Android-specific behaviors. It reconstructs these behaviors by observing and dissecting system calls and, therefore, is resistant to the multitude of alterations the Android runtime is subjected to over its life-cycle. CopperDroid automatically and accurately reconstructs events of interest that describe, not only well-known process-OS interactions (e.g., file and process creation), but also complex intra-and inter-process communications (e.g., SMS reception), whose semantics are typically contextualized through complex Android objects. Because CopperDroid's reconstruction mechanisms are agnostic to the underlying action invocation methods, it is able to capture actions initiated both from Java and native code execution. CopperDroid's analysis generates detailed behavioral profiles that abstract a large stream of low-level-often uninteresting-events into concise, high-level semantics, which are well-suited to provide insightful behavioral traits and open the possibility to further research directions. We carried out an extensive evaluation to assess the capabilities and performance of CopperDroid on more than 2,900 Android malware samples. Our experiments show that CopperDroid faithfully reconstructs OSand Android-specific behaviors. Additionally, we demonstrate how CopperDroid can be leveraged to disclose additional behaviors through the use of a simple, yet effective, app stimulation technique. Using this technique, we successfully triggered and disclosed additional behaviors on more than 60% of the analyzed malware samples. This qualitatively demonstrates the versatility of CopperDroid's ability to improve dynamic-based code coverage. Permission to freely reproduce all or part of this paper for noncommercial purposes is granted provided that copies bear this notice and the full citation on the first page. Reproduction for commercial purposes is strictly prohibited without the prior written consent of the Internet Society, the first-named author (for reproduction of an entire paper only), and the author's employer if the paper was prepared within the scope of employment.

Abstract. Modern botnets rely on domain-generation algorithms (DGAs) to build resilient command-and-control infrastructures. Given the prevalence of this mechanism, recent work has focused on the analysis of DNS traffic to recognize botnets based on their DGAs. While previous work has concentrated on detection, we focus on supporting intelligence operations. We propose Phoenix, a mechanism that, in addition to telling DGA-and non-DGA-generated domains apart using a combination of string and IP-based features, characterizes the DGAs behind them, and, most importantly, finds groups of DGA-generated domains that are representative of the respective botnets. As a result, Phoenix can associate previously unknown DGA-generated domains to these groups, and produce novel knowledge about the evolving behavior of each tracked botnet. We evaluated Phoenix on 1,153,516 domains, including DGA-generated domains from modern, well-known botnets: without supervision, it correctly distinguished DGA-vs. non-DGA-generated domains in 94.8 percent of the cases, characterized families of domains that belonged to distinct DGAs, and helped researchers "on the field" in gathering intelligence on suspicious domains to identify the correct botnet.

Abstract-The Android packaging model offers ample opportunities for malware writers to piggyback malicious code in popular apps, which can then be easily spread to a large user base. Although recent research has produced approaches and tools to identify piggybacked apps, the literature lacks a comprehensive investigation into such phenomenon. We fill this gap by 1) systematically building a large set of piggybacked and benign apps pairs, which we release to the community, 2) empirically studying the characteristics of malicious piggybacked apps in comparison with their benign counterparts, and 3) providing insights on piggybacking processes. Among several findings providing insights analysis techniques should build upon to improve the overall detection and classification accuracy of piggybacked apps, we show that piggybacking operations not only concern app code, but also extensively manipulates app resource files, largely contradicting common beliefs. We also find that piggybacking is done with little sophistication, in many cases automatically, and often via library code.