Xueyuan Han

Xueyuan Han, Zinuo Cai, Yichu Zhang et al.

The application of Transformer-based large models has achieved numerous success in recent years. However, the exponential growth in the parameters of large models introduces formidable memory challenge for edge deployment. Prior works to address this challenge mainly focus on optimizing the model structure and adopting memory swapping methods. However, the former reduces the inference accuracy, and the latter raises the inference latency. This paper introduces PIPELOAD, a novel memory-efficient pipeline execution mechanism. It reduces memory usage by incorporating dynamic memory management and minimizes inference latency by employing parallel model loading. Based on PIPELOAD mechanism, we present Hermes, a framework optimized for large model inference on edge devices. We evaluate Hermes on Transformer-based models of different sizes. Our experiments illustrate that Hermes achieves up to 4.24 X increase in inference speed and 86.7% lower memory consumption than the state-of-the-art pipeline mechanism for BERT and ViT models, 2.58 X increase in inference speed and 90.3% lower memory consumption for GPT-style models.

Zijun Cheng, Qiujian Lv, Jinyuan Liang et al.

Provenance graphs are structured audit logs that describe the history of a system's execution. Recent studies have explored a variety of techniques to analyze provenance graphs for automated host intrusion detection, focusing particularly on advanced persistent threats. Sifting through their design documents, we identify four common dimensions that drive the development of provenance-based intrusion detection systems (PIDSes): scope (can PIDSes detect modern attacks that infiltrate across application boundaries?), attack agnosticity (can PIDSes detect novel attacks without a priori knowledge of attack characteristics?), timeliness (can PIDSes efficiently monitor host systems as they run?), and attack reconstruction (can PIDSes distill attack activity from large provenance graphs so that sysadmins can easily understand and quickly respond to system intrusion?). We present KAIROS, the first PIDS that simultaneously satisfies the desiderata in all four dimensions, whereas existing approaches sacrifice at least one and struggle to achieve comparable detection performance. Kairos leverages a novel graph neural network-based encoder-decoder architecture that learns the temporal evolution of a provenance graph's structural changes to quantify the degree of anomalousness for each system event. Then, based on this fine-grained information, Kairos reconstructs attack footprints, generating compact summary graphs that accurately describe malicious activity over a stream of system audit logs. Using state-of-the-art benchmark datasets, we demonstrate that Kairos outperforms previous approaches.

Soo Yee Lim, Bogdan Stelea, Xueyuan Han et al.

Despite the wide usage of container-based cloud computing, container auditing for security analysis relies mostly on built-in host audit systems, which often lack the ability to capture high-fidelity container logs. State-of-the-art reference-monitor-based audit techniques greatly improve the quality of audit logs, but their system-wide architecture is too costly to be adapted for individual containers. Moreover, these techniques typically require extensive kernel modifications, making them difficult to deploy in practical settings. In this paper, we present saBPF (secure audit BPF), an extension of the eBPF framework capable of deploying secure system-level audit mechanisms at the container granularity. We demonstrate the practicality of saBPF in Kubernetes by designing an audit framework, an intrusion detection system, and a lightweight access control mechanism. We evaluate saBPF and show that it is comparable in performance and security guarantees to audit systems from the literature that are implemented directly in the kernel.

Xueyuan Han, Xiao Yu, Thomas Pasquier et al.

Many users implicitly assume that software can only be exploited after it is installed. However, recent supply-chain attacks demonstrate that application integrity must be ensured during installation itself. We introduce SIGL, a new tool for detecting malicious behavior during software installation. SIGL collects traces of system call activity, building a data provenance graph that it analyzes using a novel autoencoder architecture with a graph long short-term memory network (graph LSTM) for the encoder and a standard multilayer perceptron for the decoder. SIGL flags suspicious installations as well as the specific installation-time processes that are likely to be malicious. Using a test corpus of 625 malicious installers containing real-world malware, we demonstrate that SIGL has a detection accuracy of 96%, outperforming similar systems from industry and academia by up to 87% in precision and recall and 45% in accuracy. We also demonstrate that SIGL can pinpoint the processes most likely to have triggered malicious behavior, works on different audit platforms and operating systems, and is robust to training data contamination and adversarial attack. It can be used with application-specific models, even in the presence of new software versions, as well as application-agnostic meta-models that encompass a wide range of applications and installers.

Xueyuan Han, James Mickens, Ashish Gehani et al.

Host-based anomaly detectors generate alarms by inspecting audit logs for suspicious behavior. Unfortunately, evaluating these anomaly detectors is hard. There are few high-quality, publicly-available audit logs, and there are no pre-existing frameworks that enable push-button creation of realistic system traces. To make trace generation easier, we created Xanthus, an automated tool that orchestrates virtual machines to generate realistic audit logs. Using Xanthus' simple management interface, administrators select a base VM image, configure a particular tracing framework to use within that VM, and define post-launch scripts that collect and save trace data. Once data collection is finished, Xanthus creates a self-describing archive, which contains the VM, its configuration parameters, and the collected trace data. We demonstrate that Xanthus hides many of the tedious (yet subtle) orchestration tasks that humans often get wrong; Xanthus avoids mistakes that lead to non-replicable experiments.

Xueyuan Han, Thomas Pasquier, Adam Bates et al.

Advanced Persistent Threats (APTs) are difficult to detect due to their "low-and-slow" attack patterns and frequent use of zero-day exploits. We present UNICORN, an anomaly-based APT detector that effectively leverages data provenance analysis. From modeling to detection, UNICORN tailors its design specifically for the unique characteristics of APTs. Through extensive yet time-efficient graph analysis, UNICORN explores provenance graphs that provide rich contextual and historical information to identify stealthy anomalous activities without pre-defined attack signatures. Using a graph sketching technique, it summarizes long-running system execution with space efficiency to combat slow-acting attacks that take place over a long time span. UNICORN further improves its detection capability using a novel modeling approach to understand long-term behavior as the system evolves. Our evaluation shows that UNICORN outperforms an existing state-of-the-art APT detection system and detects real-life APT scenarios with high accuracy.

Thomas Pasquier, Xueyuan Han, Thomas Moyer et al.

Identifying the root cause and impact of a system intrusion remains a foundational challenge in computer security. Digital provenance provides a detailed history of the flow of information within a computing system, connecting suspicious events to their root causes. Although existing provenance-based auditing techniques provide value in forensic analysis, they assume that such analysis takes place only retrospectively. Such post-hoc analysis is insufficient for realtime security applications, moreover, even for forensic tasks, prior provenance collection systems exhibited poor performance and scalability, jeopardizing the timeliness of query responses. We present CamQuery, which provides inline, realtime provenance analysis, making it suitable for implementing security applications. CamQuery is a Linux Security Module that offers support for both userspace and in-kernel execution of analysis applications. We demonstrate the applicability of CamQuery to a variety of runtime security applications including data loss prevention, intrusion detection, and regulatory compliance. In evaluation, we demonstrate that CamQuery reduces the latency of realtime query mechanisms, while imposing minimal overheads on system execution. CamQuery thus enables the further deployment of provenance-based technologies to address central challenges in computer security.

Xueyuan Han, Thomas Pasquier, Margo Seltzer

Intrusion detection is an arms race; attackers evade intrusion detection systems by developing new attack vectors to sidestep known defense mechanisms. Provenance provides a detailed, structured history of the interactions of digital objects within a system. It is ideal for intrusion detection, because it offers a holistic, attack-vector-agnostic view of system execution. As such, provenance graph analysis fundamentally strengthens detection robustness. We discuss the opportunities and challenges associated with provenance-based intrusion detection and provide insights based on our experience building such systems.

Xueyuan Han, Thomas Pasquier, Tanvi Ranjan et al.

We present FRAPpuccino (or FRAP), a provenance-based fault detection mechanism for Platform as a Service (PaaS) users, who run many instances of an application on a large cluster of machines. FRAP models, records, and analyzes the behavior of an application and its impact on the system as a directed acyclic provenance graph. It assumes that most instances behave normally and uses their behavior to construct a model of legitimate behavior. Given a model of legitimate behavior, FRAP uses a dynamic sliding window algorithm to compare a new instance's execution to that of the model. Any instance that does not conform to the model is identified as an anomaly. We present the FRAP prototype and experimental results showing that it can accurately detect application anomalies.

Thomas Pasquier, Xueyuan Han, Mark Goldstein et al.

Data provenance describes how data came to be in its present form. It includes data sources and the transformations that have been applied to them. Data provenance has many uses, from forensics and security to aiding the reproducibility of scientific experiments. We present CamFlow, a whole-system provenance capture mechanism that integrates easily into a PaaS offering. While there have been several prior whole-system provenance systems that captured a comprehensive, systemic and ubiquitous record of a system's behavior, none have been widely adopted. They either A) impose too much overhead, B) are designed for long-outdated kernel releases and are hard to port to current systems, C) generate too much data, or D) are designed for a single system. CamFlow addresses these shortcoming by: 1) leveraging the latest kernel design advances to achieve efficiency; 2) using a self-contained, easily maintainable implementation relying on a Linux Security Module, NetFilter, and other existing kernel facilities; 3) providing a mechanism to tailor the captured provenance data to the needs of the application; and 4) making it easy to integrate provenance across distributed systems. The provenance we capture is streamed and consumed by tenant-built auditor applications. We illustrate the usability of our implementation by describing three such applications: demonstrating compliance with data regulations; performing fault/intrusion detection; and implementing data loss prevention. We also show how CamFlow can be leveraged to capture meaningful provenance without modifying existing applications.