Adi Sosnovich

Bar Weiss, Antonio Abu-Nassar, Adi Sosnovich et al.

Code review is a critical practice in software engineering, yet the growing scale and frequency of code patches in modern projects, together with the widespread adoption of AI code assistants, make manual review increasingly challenging. Identifying the types of changes within a patch, such as renames, moves, or logic modifications, can substantially improve review efficiency by enabling prioritization, filtering, and automation. However, existing LLM-based approaches to code review have largely focused on summarization and comment generation, leaving structured code reviews underexplored. In this paper, we present a systematic study of using large language models (LLMs) for taxonomy-based labeling of code changes in a code patch. We introduce a two-stage pipeline that assigns labels to diff hunks and then refines them to capture structural relationships and semantic attributes, such as rename propagation and type changes. Our approach employs few-shot prompting to produce language-agnostic and customizable labels, without the engineering overhead of traditional static-analysis pipelines. We evaluate four LLMs across multiple context configurations on a manually curated benchmark of natural and synthetic patches. Our best configuration achieves up to $84\%$ recall and $81\%$ precision, with high accuracy in extracting relational and attribute metadata. These results suggest that LLM-based labeling can effectively complement static analysis by enabling flexible, multilingual, and automation-friendly code review workflows.

Adi Sosnovich, Orna Grumberg, Gabi Nakibly

The Internet infrastructure relies entirely on open standards for its routing protocols. However, the majority of routers on the Internet are closed-source. Hence, there is no straightforward way to analyze them. Specifically, one cannot easily identify deviations of a router's routing functionality from the routing protocol's standard. Such deviations (either deliberate or inadvertent) are particularly important to identify since they may degrade the security or resiliency of the network. A model-based testing procedure is a technique that allows to systematically generate tests based on a model of the system to be tested; thereby finding deviations in the system compared to the model. However, applying such an approach to a complex multi-party routing protocol requires a prohibitively high number of tests to cover the desired functionality. We propose efficient and practical optimizations to the model-based testing procedure that are tailored to the analysis of routing protocols. These optimizations allow to devise a formal black-box method to unearth deviations in closed-source routing protocols' implementations. The method relies only on the ability to test the targeted protocol implementation and observe its output. Identification of the deviations is fully automatic. We evaluate our method against one of the complex and widely used routing protocols on the Internet -- OSPF. We search for deviations in the OSPF implementation of Cisco. Our evaluation identified numerous significant deviations that can be abused to compromise the security of a network. The deviations were confirmed by Cisco. We further employed our method to analyze the OSPF implementation of the Quagga Routing Suite. The analysis revealed one significant deviation. Subsequent to the disclosure of the deviations some of them were also identified by IBM, Lenovo and Huawei in their own products.