Michalis Polychronakis

Muhammad Farrukh, Baris Coskun, Tapti Palit et al.

Rust is a strong contender for a memory-safe alternative to C as a "systems" language, but porting the vast amount of existing C code to Rust remains daunting. In this paper, we evaluate the potential of large language models (LLMs) to automate the transpilation of C code to idiomatic Rust. We present SafeTrans, a generic framework that leverages LLMs to i) transpile C code into Rust, and ii) iteratively repair compilation and runtime errors. A key novelty of our approach is a few-shot guided repair technique for translation errors, which provides contextual information and example code snippets for specific error types, guiding the LLM toward the correct solution. Another novel aspect of our work is the evaluation of the security implications of the transpilation process, showing how some vulnerability classes in C persist in the translated Rust code. SafeTrans was evaluated with six leading LLMs on 2,653 C programs and two real-world C projects. Our results show that iterative repair improves the rate of successful translations from 54% to 80% for the best-performing LLM (gpt-4o).

Muhammad Farrukh, Baris Coskun, Tapti Palit et al.

Large-scale migration of legacy C code to Rust offers a promising path toward improving memory safety, but LLM-based C-to-Rust translation remains challenging due to limited context windows and hallucinations. Prior approaches are evaluated primarily on small programs or datasets skewed toward small codebases, providing limited insight into scalability on real-world systems. They also rely on static context construction, which breaks down in the presence of complex cross-module dependencies and often requires manual intervention. Recent coding agents offer a promising alternative through dynamic codebase navigation and context curation. When used out of the box, however, they frequently produce incomplete translations that appear superficially correct. We present ORBIT, an autonomous agentic framework for project-level C-to-Rust translation that combines dynamic context collection with dependency-guided orchestration and iterative verification. ORBIT constructs a dependency-aware translation graph, generates Rust interfaces, maps C functions to Rust counterparts, and coordinates multiple specialized agents. We evaluate ORBIT on 24 programs from CRUST-Bench, with 91.7% of the programs exceeding 1,000 lines of code. ORBIT achieves 100% compilation success and 91.7% test success in both expert-interface and automatically generated-interface settings, substantially outperforming C2Rust and CRUST-Bench, while reducing unsafe Rust code blocks to nearly zero. We further evaluate ORBIT on challenging cases from the DARPA TRACTOR benchmark, where it achieves competitive performance relative to participating systems.

Jiyong Uhm, Minseok Kim, Michalis Polychronakis et al.

Binary code analysis plays an essential role in cybersecurity, facilitating reverse engineering to reveal the inner workings of programs in the absence of source code. Traditional approaches, such as static and dynamic analysis, extract valuable insights from stripped binaries, but often demand substantial expertise and manual effort. Recent advances in deep learning have opened promising opportunities to enhance binary analysis by capturing latent features and disclosing underlying code semantics. Despite the growing number of binary analysis models based on machine learning, their robustness to adversarial code transformations at the binary level remains underexplored. We evaluate the robustness of deep learning models for the task of binary code similarity detection (BCSD) under semantics-preserving transformations. The unique nature of machine instructions presents distinct challenges compared to the typical input perturbations found in other domains. We introduce asmFooler, a system that evaluates the resilience of BCSD models using a diverse set of adversarial code transformations that preserve functional semantics. We construct a dataset of 9,565 binary variants from 620 baseline samples by applying eight semantics-preserving transformations across six representative BCSD models. Our major findings highlight several key insights: i) model robustness relies on the processing pipeline, including code pre-processing, architecture, and feature selection; ii) adversarial transformation effectiveness is bounded by a budget shaped by model-specific constraints like input size and instruction expressive capacity; iii) well-crafted transformations can be highly effective with minimal perturbations; and iv) such transformations efficiently disrupt model decisions (e.g., misleading to false positives or false negatives) by focusing on semantically significant instructions.

Nguyen Phong Hoang, Michalis Polychronakis, Phillipa Gill

Most online communications rely on DNS to map domain names to their hosting IP address(es). Previous work has shown that DNS-based network interference is widespread due to the unencrypted and unauthenticated nature of the original DNS protocol. In addition to DNS, accessed domain names can also be monitored by on-path observers during the TLS handshake when the SNI extension is used. These lingering issues with exposed plaintext domain names have led to the development of a new generation of protocols that keep accessed domain names hidden. DNS-over-TLS (DoT) and DNS-over-HTTPS (DoH) hide the domain names of DNS queries, while Encrypted Server Name Indication (ESNI) encrypts the domain name in the SNI extension. We present DNEye, a measurement system built on top of a network of distributed vantage points, which we used to study the accessibility of DoT/DoH and ESNI, and to investigate whether these protocols are tampered with by network providers (e.g., for censorship). Moreover, we evaluate the efficacy of these protocols in circumventing network interference when accessing content blocked by traditional DNS manipulation. We find evidence of blocking efforts against domain name encryption technologies in several countries, including China, Russia, and Saudi Arabia. At the same time, we discover that domain name encryption can help with unblocking more than 55% and 95% of censored domains in China and other countries where DNS-based filtering is heavily employed.

Nguyen Phong Hoang, Arian Akhavan Niaki, Jakub Dalek et al.

The DNS filtering apparatus of China's Great Firewall (GFW) has evolved considerably over the past two decades. However, most prior studies of China's DNS filtering were performed over short time periods, leading to unnoticed changes in the GFW's behavior. In this study, we introduce GFWatch, a large-scale, longitudinal measurement platform capable of testing hundreds of millions of domains daily, enabling continuous monitoring of the GFW's DNS filtering behavior. We present the results of running GFWatch over a nine-month period, during which we tested an average of 411M domains per day and detected a total of 311K domains censored by GFW's DNS filter. To the best of our knowledge, this is the largest number of domains tested and censored domains discovered in the literature. We further reverse engineer regular expressions used by the GFW and find 41K innocuous domains that match these filters, resulting in overblocking of their content. We also observe bogus IPv6 and globally routable IPv4 addresses injected by the GFW, including addresses owned by US companies, such as Facebook, Dropbox, and Twitter. Using data from GFWatch, we studied the impact of GFW blocking on the global DNS system. We found 77K censored domains with DNS resource records polluted in popular public DNS resolvers, such as Google and Cloudflare. Finally, we propose strategies to detect poisoned responses that can (1) sanitize poisoned DNS records from the cache of public DNS resolvers, and (2) assist in the development of circumvention tools to bypass the GFW's DNS censorship.

Nguyen Phong Hoang, Arian Akhavan Niaki, Phillipa Gill et al.

Although the security benefits of domain name encryption technologies such as DNS over TLS (DoT), DNS over HTTPS (DoH), and Encrypted Client Hello (ECH) are clear, their positive impact on user privacy is weakened by--the still exposed--IP address information. However, content delivery networks, DNS-based load balancing, co-hosting of different websites on the same server, and IP address churn, all contribute towards making domain-IP mappings unstable, and prevent straightforward IP-based browsing tracking. In this paper, we show that this instability is not a roadblock (assuming a universal DoT/DoH and ECH deployment), by introducing an IP-based website fingerprinting technique that allows a network-level observer to identify at scale the website a user visits. Our technique exploits the complex structure of most websites, which load resources from several domains besides their primary one. Using the generated fingerprints of more than 200K websites studied, we could successfully identify 84% of them when observing solely destination IP addresses. The accuracy rate increases to 92% for popular websites, and 95% for popular and sensitive websites. We also evaluated the robustness of the generated fingerprints over time, and demonstrate that they are still effective at successfully identifying about 70% of the tested websites after two months. We conclude by discussing strategies for website owners and hosting providers towards hindering IP-based website fingerprinting and maximizing the privacy benefits offered by DoT/DoH and ECH.

Nguyen Phong Hoang, Arian Akhavan Niaki, Michalis Polychronakis et al.

Understanding web co-location is essential for various reasons. For instance, it can help one to assess the collateral damage that denial-of-service attacks or IP-based blocking can cause to the availability of co-located web sites. However, it has been more than a decade since the first study was conducted in 2007. The Internet infrastructure has changed drastically since then, necessitating a renewed study to comprehend the nature of web co-location. In this paper, we conduct an empirical study to revisit web co-location using datasets collected from active DNS measurements. Our results show that the web is still small and centralized to a handful of hosting providers. More specifically, we find that more than 60% of web sites are co-located with at least ten other web sites---a group comprising less popular web sites. In contrast, 17.5% of mostly popular web sites are served from their own servers. Although a high degree of web co-location could make co-hosted sites vulnerable to DoS attacks, our findings show that it is an increasing trend to co-host many web sites and serve them from well-provisioned content delivery networks (CDN) of major providers that provide advanced DoS protection benefits. Regardless of the high degree of web co-location, our analyses of popular block lists indicate that IP-based blocking does not cause severe collateral damage as previously thought.

Nguyen Phong Hoang, Ivan Lin, Seyedhamed Ghavamnia et al.

Centralized DNS over HTTPS/TLS (DoH/DoT) resolution, which has started being deployed by major hosting providers and web browsers, has sparked controversy among Internet activists and privacy advocates due to several privacy concerns. This design decision causes the trace of all DNS resolutions to be exposed to a third-party resolver, different than the one specified by the user's access network. In this work we propose K-resolver, a DNS resolution mechanism that disperses DNS queries across multiple DoH resolvers, reducing the amount of information about a user's browsing activity exposed to each individual resolver. As a result, none of the resolvers can learn a user's entire web browsing history. We have implemented a prototype of our approach for Mozilla Firefox, and used it to evaluate the performance of web page load time compared to the default centralized DoH approach. While our K-resolver mechanism has some effect on DNS resolution time and web page load time, we show that this is mainly due to the geographical location of the selected DoH servers. When more well-provisioned anycast servers are available, our approach incurs negligible overhead while improving user privacy.

Nguyen Phong Hoang, Arian Akhavan Niaki, Nikita Borisov et al.

As Internet users have become more savvy about the potential for their Internet communication to be observed, the use of network traffic encryption technologies (e.g., HTTPS/TLS) is on the rise. However, even when encryption is enabled, users leak information about the domains they visit via DNS queries and via the Server Name Indication (SNI) extension of TLS. Two recent proposals to ameliorate this issue are DNS over HTTPS/TLS (DoH/DoT) and Encrypted SNI (ESNI). In this paper we aim to assess the privacy benefits of these proposals by considering the relationship between hostnames and IP addresses, the latter of which are still exposed. We perform DNS queries from nine vantage points around the globe to characterize this relationship. We quantify the privacy gain offered by ESNI for different hosting and CDN providers using two different metrics, the k-anonymity degree due to co-hosting and the dynamics of IP address changes. We find that 20% of the domains studied will not gain any privacy benefit since they have a one-to-one mapping between their hostname and IP address. On the other hand, 30% will gain a significant privacy benefit with a k value greater than 100, since these domains are co-hosted with more than 100 other domains. Domains whose visitors' privacy will meaningfully improve are far less popular, while for popular domains the benefit is not significant. Analyzing the dynamics of IP addresses of long-lived domains, we find that only 7.7% of them change their hosting IP addresses on a daily basis. We conclude by discussing potential approaches for website owners and hosting/CDN providers for maximizing the privacy benefits of ESNI.

Panagiotis Papadopoulos, Panagiotis Ilia, Michalis Polychronakis et al.

The proliferation of web applications has essentially transformed modern browsers into small but powerful operating systems. Upon visiting a website, user devices run implicitly trusted script code, the execution of which is confined within the browser to prevent any interference with the user's system. Recent JavaScript APIs, however, provide advanced capabilities that not only enable feature-rich web applications, but also allow attackers to perform malicious operations despite the confined nature of JavaScript code execution. In this paper, we demonstrate the powerful capabilities that modern browser APIs provide to attackers by presenting MarioNet: a framework that allows a remote malicious entity to control a visitor's browser and abuse its resources for unwanted computation or harmful operations, such as cryptocurrency mining, password-cracking, and DDoS. MarioNet relies solely on already available HTML5 APIs, without requiring the installation of any additional software. In contrast to previous browser-based botnets, the persistence and stealthiness characteristics of MarioNet allow the malicious computations to continue in the background of the browser even after the user closes the window or tab of the initial malicious website. We present the design, implementation, and evaluation of a prototype system, MarioNet, that is compatible with all major browsers, and discuss potential defense strategies to counter the threat of such persistent in-browser attacks. Our main goal is to raise awareness regarding this new class of attacks, and inform the design of future browser APIs so that they provide a more secure client-side environment for web applications.