CharBot: A Simple and Effective Method for Evading DGA Classifiers

Peck, Jonathan; Nie, Claire; Sivaguru, Raaghavi; Grumer, Charles; Olumofin, Femi; Yu, Bin; Nascimento, Anderson C. A.; Cock, Martine De

doi:10.48550/arxiv.1905.01078

Cited by 2 publications

(10 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The effectiveness of the adversarial samples generated by these strategies have so far only been evaluated against classifiers trained on resolving DNS traffic. We show that, when trained on NX-traffic, the classifiers are more robust and correctly classify the adversarial samples generated by the strategies described in [15,25].…”

Section: Related Workmentioning

confidence: 94%

“…Recently, some of the DGA classifiers have been shown to be vulnerable to adversarial examples which are worst-case perturbations of input data causing a classifier to misclassify such samples (e.g. [1,15,22,25]). In [25] the samples are generated by exploiting known features of a feature-based classifier.…”

Section: Related Workmentioning

confidence: 99%

“…In [25] the samples are generated by exploiting known features of a feature-based classifier. In [15] they are created by slightly perturbing Alexa top domain names, exploiting the fact that many of the classifiers are trained using resolving DNS traffic. In [1] a generative adversarial network is leveraged to generate adversarial samples.…”

Section: Related Workmentioning

confidence: 99%

“…Peck et al propose Charbot [15], a DGA which randomly selects a second level domain from the Alexa top domain names, replaces two random characters and appends one of 22 TLDs in order to generate an AGD. Spooren et al iteratively developed the Deception DGA [25] targeting FANCI by exploiting the handcrafted features used for classification.…”

Section: Robustnessmentioning

confidence: 99%

“…Recall that in our setup all classifiers are trained on real-world NXDs as benign samples. In comparison, Table 7 also includes the TPRs of FANCI and B-Endgame reported in [15], using resolving DNS traffic or Alexa top domain names as source for benign samples. Our evaluation shows that all classifiers are remarkably robust against both adversarial attacks when NX-traffic is used as benign data for training.…”

Section: Robustnessmentioning

confidence: 99%

See 4 more Smart Citations

Analyzing the real-world applicability of DGA classifiers

Drichel

Meyer

Schüppen

et al. 2020

Proceedings of the 15th International Conference on Availability, Reliability and Security

View full text Add to dashboard Cite

Separating benign domains from domains generated by DGAs with the help of a binary classifier is a well-studied problem for which promising performance results have been published. The corresponding multiclass task of determining the exact DGA that generated a domain enabling targeted remediation measures is less well studied. Selecting the most promising classifier for these tasks in practice raises a number of questions that have not been addressed in prior work so far. These include the questions on which traffic to train in which network and when, just as well as how to assess robustness against adversarial attacks. Moreover, it is unclear which features lead a classifier to a decision and whether the classifiers are real-time capable. In this paper, we address these issues and thus contribute to bringing DGA detection classifiers closer to practical use. In this context, we propose one novel classifier based on residual neural networks for each of the two tasks and extensively evaluate them as well as previously proposed classifiers in a unified setting. We not only evaluate their classification performance but also compare them with respect to explainability, robustness, and training and classification speed. Finally, we show that our newly proposed binary classifier generalizes well to other networks, is time-robust, and able to identify previously unknown DGAs. CCS CONCEPTS• Security and privacy → Intrusion detection systems; Malware and its mitigation; • Computing methodologies → Machine learning.

show abstract

Section: Related Workmentioning

confidence: 94%