Biblio

Deep learning technologies, which are the key components of state-of-the-art Artificial Intelligence (AI) services, have shown great success in providing human-level capabilities for a variety of tasks, such as visual analysis, speech recognition, and natural language processing and etc. Building a production-level deep learning model is a non-trivial task, which requires a large amount of training data, powerful computing resources, and human expertises. Therefore, illegitimate reproducing, distribution, and the derivation of proprietary deep learning models can lead to copyright infringement and economic harm to model creators. Therefore, it is essential to devise a technique to protect the intellectual property of deep learning models and enable external verification of the model ownership. In this paper, we generalize the "digital watermarking'' concept from multimedia ownership verification to deep neural network (DNNs) models. We investigate three DNN-applicable watermark generation algorithms, propose a watermark implanting approach to infuse watermark into deep learning models, and design a remote verification mechanism to determine the model ownership. By extending the intrinsic generalization and memorization capabilities of deep neural networks, we enable the models to learn specially crafted watermarks at training and activate with pre-specified predictions when observing the watermark patterns at inference. We evaluate our approach with two image recognition benchmark datasets. Our framework accurately (100$\backslash$%) and quickly verifies the ownership of all the remotely deployed deep learning models without affecting the model accuracy for normal input data. In addition, the embedded watermarks in DNN models are robust and resilient to different counter-watermark mechanisms, such as fine-tuning, parameter pruning, and model inversion attacks.

Advanced cyber attacks consist of multiple stages aimed at being stealthy and elusive. Such attack patterns leave their footprints spatio-temporally dispersed across many different logs in victim machines. However, existing log-mining intrusion analysis systems typically target only a single type of log to discover evidence of an attack and therefore fail to exploit fundamental inter-log connections. The output of such single-log analysis can hardly reveal the complete attack story for complex, multi-stage attacks. Additionally, some existing approaches require heavyweight system instrumentation, which makes them impractical to deploy in real production environments. To address these problems, we present HERCULE, an automated multi-stage log-based intrusion analysis system. Inspired by graph analytics research in social network analysis, we model multi-stage intrusion analysis as a community discovery problem. HERCULE builds multi-dimensional weighted graphs by correlating log entries across multiple lightweight logs that are readily available on commodity systems. From these, HERCULE discovers any "attack communities" embedded within the graphs. Our evaluation with 15 well known APT attack families demonstrates that HERCULE can reconstruct attack behaviors from a spectrum of cyber attacks that involve multiple stages with high accuracy and low false positive rates.