Biblio

We have recently seen an increasing number of attacks that are distributed, and span an entire wide area network (WAN). Today, typically, intrusion detection systems (IDSs) are deployed at enterprise scale and cannot handle attacks that cover a WAN. Moreover, such IDSs are implemented at a single entity that expects to look at all packets to determine an intrusion. Transferring copies of raw packets to centralized engines for analysis in a WAN can significantly impact both network performance and detection accuracy. In this paper, we propose Jaal, a framework for achieving accurate network intrusion detection at scale. The key idea in Jaal is to monitor traffic and construct in-network packet summaries. The summaries are then processed centrally to detect attacks with high accuracy. The main challenges that we address are (a) creating summaries that are concise, but sufficient to draw highly accurate inferences and (b) transforming traditional IDS rules to handle summaries instead of raw packets. We implement Jaal on a large scale SDN testbed. We show that on average Jaal yields a detection accuracy of about 98%, which is the highest reported for ISP scale network intrusion detection. At the same time, the overhead associated with transferring summaries to the central inference engine is only about 35% of what is consumed if raw packets are transferred.

Machine learning (ML) models, e.g., deep neural networks (DNNs), are vulnerable to adversarial examples: malicious inputs modified to yield erroneous model outputs, while appearing unmodified to human observers. Potential attacks include having malicious content like malware identified as legitimate or controlling vehicle behavior. Yet, all existing adversarial example attacks require knowledge of either the model internals or its training data. We introduce the first practical demonstration of an attacker controlling a remotely hosted DNN with no such knowledge. Indeed, the only capability of our black-box adversary is to observe labels given by the DNN to chosen inputs. Our attack strategy consists in training a local model to substitute for the target DNN, using inputs synthetically generated by an adversary and labeled by the target DNN. We use the local substitute to craft adversarial examples, and find that they are misclassified by the targeted DNN. To perform a real-world and properly-blinded evaluation, we attack a DNN hosted by MetaMind, an online deep learning API. We find that their DNN misclassifies 84.24% of the adversarial examples crafted with our substitute. We demonstrate the general applicability of our strategy to many ML techniques by conducting the same attack against models hosted by Amazon and Google, using logistic regression substitutes. They yield adversarial examples misclassified by Amazon and Google at rates of 96.19% and 88.94%. We also find that this black-box attack strategy is capable of evading defense strategies previously found to make adversarial example crafting harder.