The binary similarity problem consists in determining if two functions are similar considering only their compiled form. Advanced techniques for binary similarity recently gained momentum as they can be applied in several fields, such as copyright disputes, malware analysis, vulnerability detection, etc. In this article we describe SAFE, a novel architecture for function representation based on a self-attentive neural network. SAFE works directly on disassembled binary functions, does not require manual feature extraction, is computationally more efficient than existing solutions, and is more general as it works on stripped binaries and on multiple architectures. Results from our experimental evaluation show how SAFE provides a performance improvement with respect to previous solutions. Furthermore, we show how SAFE can be used in widely different use cases, thus providing a general solution for several application scenarios.
The vast majority of today’s mobile malware targets Android devices. An important task of malware analysis is the classification of malicious samples into known families. In this paper, we propose AndroDFA (DFA, detrended fluctuation analysis): an approach to Android malware family classification based on dynamic analysis of resource consumption metrics available from the proc file system. These metrics can be easily measured during sample execution. From each malware, we extract features through detrended fluctuation analysis (DFA) and Pearson’s correlation, then a support vector machine is employed to classify malware into families. We provide an experimental evaluation based on malware samples from two datasets, namely Drebin and AMD. With the Drebin dataset, we obtained a classification accuracy of 82%, comparable with works from the state-of-the-art like DroidScribe. However, compared to DroidScribe, our approach is easier to reproduce because it is based on publicly available tools only, does not require any modification to the emulated environment or Android OS, and by design, can also be used on physical devices rather than exclusively on emulators. The latter is a key factor because modern mobile malware can detect the emulated environment and hide its malicious behavior. The experiments on the AMD dataset gave similar results, with an overall mean accuracy of 78%. Furthermore, we made the software we developed publicly available, to ease the reproducibility of our results.
Embedded systems are resource-constrained and therefore cannot sustain running advanced security software on-board, which makes them an ideal target for today’s malware. This calls for advancements in malware analysis techniques that can be applied to embedded systems. An important task of malware analysis is the classification of malicious samples into known families, to extract valuable intelligence about the main features of a malware without the need to dissect it. Classifying malware in families is particularly important when it comes to analyzing malicious software in platforms that are massively targeted, such as Android, because it helps reduce the workload within the analysis pipeline. In this paper, we propose AndroDFA (DFA, detrended fluctuation analysis) : an approach to Android malware family classification based on dynamic analysis of resource consumption metrics available from the proc file system. These metrics can be easily measured during sample execution. From each malware, we extract features through detrended fluctuation analysis (DFA) and Pearson’s correlation, then use a support vector machine to classify malware into families. We provide an experimental evaluation based on malware samples from two datasets, namely Drebin and AMD . With the Drebin dataset, we obtained a classification accuracy of 82%, comparable with works from the state-of-the-art like DroidScribe. However, compared to DroidScribe, our approach is easier to reproduce because it is based on publicly available tools only, does not require any modification to the emulated environment or Android OS, and by design, can also be used on physical devices rather than exclusively on emulators. The latter is a key factor because modern mobile malware can detect the emulated environment and hide its malicious behavior. The experiments on the AMD dataset gave similar results, with an overall mean accuracy of 78%. Furthermore, we made the software we developed publicly available, to ease the reproducibility of our results.
Recent progress in pre-trained language models led to systems that are able to generate text of an increasingly high quality. While several works have investigated the fluency and grammatical correctness of such models, it is still unclear to which extent the generated text is consistent with factual world knowledge. Here, we go beyond fluency and also investigate the verifiability of text generated by state-of-the-art pre-trained language models. A generated sentence is verifiable if it can be corroborated or disproved by Wikipedia, and we find that the verifiability of generated text strongly depends on the decoding strategy. In particular, we discover a tradeoff between factuality (i.e., the ability of generating Wikipedia corroborated text) and repetitiveness. While decoding strategies such as top-k and nucleus sampling lead to less repetitive generations, they also produce less verifiable text. Based on these finding, we introduce a simple and effective decoding strategy which, in comparison to previously used decoding strategies, produces less repetitive and more verifiable text.
In this paper we investigate the problem of automatically naming pieces of assembly code. Where by naming we mean assigning to an assembly function a string of words that would likely be assigned by a human reverse engineer. We formally and precisely define the framework in which our investigation takes place. That is we define the problem, we provide reasonable justifications for the choices that we made for the design of training and the tests. We performed an analysis on a large real-world corpora constituted by nearly 9 millions of functions taken from more than 22k softwares. In such framework we test baselines coming from the field of Natural Language Processing (e.g., Seq2Seq networks and Transformer). Interestingly, our evaluation shows promising results beating the state-of-the-art and reaching good performance. We investigate the applicability of tine-tuning (i.e., taking a model already trained on a large generic corpora and retraining it for a specific task). Such technique is popular and well-known in the NLP field. Our results confirm that fine-tuning is effective even when neural networks are applied to binaries. We show that a model, pre-trained on the aforementioned corpora, when fine-tuned has higher performances on specific domains (such as predicting names in system utilites, malware, etc).