Security and Artificial Intelligence 

About: Security and Artificial Intelligence is an academic conference. The conference publishes majorly in the area(s): Computer science & Malware. Over the lifetime, 51 publications have been published by the conference receiving 2265 citations.

Adversarial machine learning

Abstract: In this paper (expanded from an invited talk at AISEC 2010), we discuss an emerging field of study: adversarial machine learning---the study of effective machine learning techniques against an adversarial opponent. In this paper, we: give a taxonomy for classifying attacks against online machine learning algorithms; discuss application-specific factors that limit an adversary's capabilities; introduce two models for modeling an adversary's capabilities; explore the limits of an adversary's knowledge about the algorithm, feature space, training, and input data; explore vulnerabilities in machine learning algorithms; discuss countermeasures against attacks; introduce the evasion challenge; and discuss privacy-preserving learning techniques.

A comparative assessment of malware classification using binary texture analysis and dynamic analysis

Abstract: AI techniques play an important role in automated malware classification. Several machine-learning methods have been applied to classify or cluster malware into families, based on different features derived from dynamic review of the malware. While these approaches demonstrate promise, they are themselves subject to a growing array of counter measures that increase the cost of capturing these binary features. Further, feature extraction requires a time investment per binary that does not scale well to the daily volume of binary instances being reported by those who diligently collect malware. Recently, a new type of feature extraction, used by a classification approach called binary-texture analysis, was introduced in [16]. We compare this approach to existing malware classification approaches previously published. We find that, while binary texture analysis is capable of providing comparable classification accuracy to that of contemporary dynamic techniques, it can deliver these results 4000 times faster than dynamic techniques. Also surprisingly, the texture-based approach seems resilient to contemporary packing strategies, and can robustly classify a large corpus of malware with both packed and unpacked samples. We present our experimental results from three independent malware corpora, comprised of over 100 thousand malware samples. These results suggest that binary-texture analysis could be a useful and efficient complement to dynamic analysis.

Improving malware classification: bridging the static/dynamic gap

Abstract: Malware classification systems have typically used some machine learning algorithm in conjunction with either static or dynamic features collected from the binary. Recently, more advanced malware has introduced mechanisms to avoid detection in these views by using obfuscation techniques to avoid static detection and execution-stalling techniques to avoid dynamic detection. In this paper we construct a classification framework that is able to incorporate both static and dynamic views into a unified framework in the hopes that, while a malicious executable can disguise itself in some views, disguising itself in every view while maintaining malicious intent will prove to be substantially more difficult. Our method uses kernels to place a similarity metric on each distinct view and then employs multiple kernel learning to find a weighted combination of the data sources which yields the best classification accuracy in a support vector machine classifier. Our approach opens up new avenues of malware research which will allow the research community to elegantly look at multiple facets of malware simultaneously, and which can easily be extended to integrate any new data sources that may become popular in the future.

Using spatio-temporal information in API calls with machine learning algorithms for malware detection

Abstract: Run-time monitoring of program execution behavior is widely used to discriminate between benign and malicious processes running on an end-host. Towards this end, most of the existing run-time intrusion or malware detection techniques utilize information available in Windows Application Programming Interface (API) call arguments or sequences. In comparison, the key novelty of our proposed tool is the use of statistical features which are extracted from both spatial arguments) and temporal (sequences) information available in Windows API calls. We provide this composite feature set as an input to standard machine learning algorithms to raise the final alarm. The results of our experiments show that the concurrent analysis of spatio-temporal features improves the detection accuracy of all classifiers. We also perform the scalability analysis to identify a minimal subset of API categories to be monitored whilst maintaining high detection accuracy.

Active learning for network intrusion detection

Abstract: Anomaly detection for network intrusion detection is usually considered an unsupervised task. Prominent techniques, such as one-class support vector machines, learn a hypersphere enclosing network data, mapped to a vector space, such that points outside of the ball are considered anomalous. However, this setup ignores relevant information such as expert and background knowledge. In this paper, we rephrase anomaly detection as an active learning task. We propose an effective active learning strategy to query low-confidence observations and to expand the data basis with minimal labeling effort. Our empirical evaluation on network intrusion detection shows that our approach consistently outperforms existing methods in relevant scenarios.