arXiv subjects

Mira Mezini

Publications and source records attributed to Mira Mezini.

At least 37 records · Page 2Linked to original sources

arXiv2025

Problem Solving Through Human-AI Preference-Based Cooperation

While there is a widespread belief that artificial general intelligence (AGI) -- or even superhuman AI -- is imminent, complex problems in expert domains are far from being solved. We argue that such problems require human-AI cooperation and that the current state of the art in generative AI is unable to play the role of a reliable partner due to a multitude of shortcomings, including difficulty to keep track of a complex solution artifact (e.g., a software program), limited support for versatile human preference expression and lack of adapting to human preference in an interactive setting. To address these challenges, we propose HAICo2, a novel human-AI co-construction framework. We take first steps towards a formalization of HAICo2 and discuss the difficult open research problems that it faces.

cs.AI↗

arXiv2025

Scaling Up: Revisiting Mining Android Sandboxes at Scale for Malware Classification

The widespread use of smartphones in daily life has raised concerns about privacy and security among researchers and practitioners. Privacy issues are generally highly prevalent in mobile applications, particularly targeting the Android platform, the most popular mobile operating system. For this reason, several techniques have been proposed to identify malicious behavior in Android applications, including the Mining Android Sandbox approach (MAS approach), which aims to identify malicious behavior in repackaged Android applications (apps). However, previous empirical studies evaluated the MAS approach using a small dataset consisting of only 102 pairs of original and repackaged apps. This limitation raises questions about the external validity of their findings and whether the MAS approach can be generalized to larger datasets. To address these concerns, this paper presents the results of a replication study focused on evaluating the performance of the MAS approach regarding its capabilities of correctly classifying malware from different families. Unlike previous studies, our research employs a dataset that is an order of magnitude larger, comprising 4,076 pairs of apps covering a more diverse range of Android malware families. Surprisingly, our findings indicate a poor performance of the MAS approach for identifying malware, with the F1-score decreasing from 0.90 for the small dataset used in the previous studies to 0.54 in our more extensive dataset. Upon closer examination, we discovered that certain malware families partially account for the low accuracy of the MAS approach, which fails to classify a repackaged version of an app as malware correctly. Our findings highlight the limitations of the MAS approach, particularly when scaled, and underscore the importance of complementing it with other techniques to detect a broader range of malware effectively.

cs.CR↗

arXiv2025

Integrating Symbolic Execution into the Fine-Tuning of Code-Generating LLMs

Code-generating Large Language Models (LLMs) have become essential tools in modern software development, enhancing productivity and accelerating development. This paper aims to investigate the fine-tuning of code-generating LLMs using Reinforcement Learning and Direct Preference Optimization, further improving their performance. To achieve this, we enhance the training data for the reward model with the help of symbolic execution techniques, ensuring more comprehensive and objective data. With symbolic execution, we create a custom dataset that better captures the nuances in code evaluation. Our reward models, fine-tuned on this dataset, demonstrate significant improvements over the baseline, CodeRL, in estimating the quality of generated code. Our code-generating LLMs, trained with the help of reward model feedback, achieve similar results compared to the CodeRL benchmark.

cs.SE↗

arXiv2025

Neural-Guided Equation Discovery

Deep learning approaches are becoming increasingly attractive for equation discovery. We show the advantages and disadvantages of using neural-guided equation discovery by giving an overview of recent papers and the results of experiments using our modular equation discovery system MGMT ($\textbf{M}$ulti-Task $\textbf{G}$rammar-Guided $\textbf{M}$onte-Carlo $\textbf{T}$ree Search for Equation Discovery). The system uses neural-guided Monte-Carlo Tree Search (MCTS) and supports both supervised and reinforcement learning, with a search space defined by a context-free grammar. We summarize seven desirable properties of equation discovery systems, emphasizing the importance of embedding tabular data sets for such learning approaches. Using the modular structure of MGMT, we compare seven architectures (among them, RNNs, CNNs, and Transformers) for embedding tabular datasets on the auxiliary task of contrastive learning for tabular data sets on an equation discovery task. For almost all combinations of modules, supervised learning outperforms reinforcement learning. Moreover, our experiments indicate an advantage of using grammar rules as action space instead of tokens. Two adaptations of MCTS -- risk-seeking MCTS and AmEx-MCTS -- can improve equation discovery with that kind of search.

cs.AI↗

arXiv2024

A Critical Study of What Code-LLMs (Do Not) Learn

Large Language Models trained on code corpora (code-LLMs) have demonstrated impressive performance in various coding assistance tasks. However, despite their increased size and training dataset, code-LLMs still have limitations such as suggesting codes with syntactic errors, variable misuse etc. Some studies argue that code-LLMs perform well on coding tasks because they use self-attention and hidden representations to encode relations among input tokens. However, previous works have not studied what code properties are not encoded by code-LLMs. In this paper, we conduct a fine-grained analysis of attention maps and hidden representations of code-LLMs. Our study indicates that code-LLMs only encode relations among specific subsets of input tokens. Specifically, by categorizing input tokens into syntactic tokens and identifiers, we found that models encode relations among syntactic tokens and among identifiers, but they fail to encode relations between syntactic tokens and identifiers. We also found that fine-tuned models encode these relations poorly compared to their pre-trained counterparts. Additionally, larger models with billions of parameters encode significantly less information about code than models with only a few hundred million parameters.

cs.SE↗

arXiv2024

Compiling with Arrays

Linear algebra computations are foundational for neural networks and machine learning, often handled through arrays. While many functional programming languages feature lists and recursion, arrays in linear algebra demand constant-time access and bulk operations. To bridge this gap, some languages represent arrays as (eager) functions instead of lists. In this paper, we connect this idea to a formal logical foundation by interpreting functions as the usual negative types from polarized type theory, and arrays as the corresponding dual positive version of the function type. Positive types are defined to have a single elimination form whose computational interpretation is pattern matching. Just like (positive) product types bind two variables during pattern matching, (positive) array types bind variables with multiplicity during pattern matching. We follow a similar approach for Booleans by introducing conditionally-defined variables. The positive formulation for the array type enables us to combine typed partial evaluation and common subexpression elimination into an elegant algorithm whose result enjoys a property we call maximal fission, which we argue can be beneficial for further optimizations. For this purpose, we present the novel intermediate representation indexed administrative normal form (AiNF), which relies on the formal logical foundation of the positive formulation for the array type to facilitate maximal loop fission and subsequent optimizations. AiNF is normal with regard to commuting conversion for both let-bindings and for-loops, leading to flat and maximally fissioned terms. We mechanize the translation and normalization from a simple surface language to AiNF, establishing that the process terminates, preserves types, and produces maximally fissioned terms.

cs.PL↗

arXiv2024

Distributed Locking as a Data Type

Mixed-consistency programming models assist programmers in designing applications that provide high availability while still ensuring application-specific safety invariants. However, existing models often make specific system assumptions, such as building on a particular database system or having baked-in coordination strategies. This makes it difficult to apply these strategies in diverse settings, ranging from client/server to ad-hoc peer-to-peer networks. This work proposes a new strategy for building programmable coordination mechanisms based on the algebraic replicated data types (ARDTs) approach. ARDTs allow for simple and composable implementations of various protocols, while making minimal assumptions about the network environment. As a case study, two different locking protocols are presented, both implemented as ARDTs. In addition, we elaborate on our ongoing efforts to integrate the approach into the LoRe mixed-consistency programming language.

cs.PL↗

arXiv2024

Supporting Error Chains in Static Analysis for Precise Evaluation Results and Enhanced Usability

Context: Static analyses are well-established to aid in understanding bugs or vulnerabilities during the development process or in large-scale studies. A low false-positive rate is essential for the adaption in practice and for precise results of empirical studies. Unfortunately, static analyses tend to report where a vulnerability manifests rather than the fix location. This can cause presumed false positives or imprecise results. Method: To address this problem, we designed an adaption of an existing static analysis algorithm that can distinguish between a manifestation and fix location, and reports error chains. An error chain represents at least two interconnected errors that occur successively, thus building the connection between the fix and manifestation location. We used our tool CogniCryptSUBS for a case study on 471 GitHub repositories, a performance benchmark to compare different analysis configurations, and conducted an expert interview. Result: We found that 50 % of the projects with a report had at least one error chain. Our runtime benchmark demonstrated that our improvement caused only a minimal runtime overhead of less than 4 %. The results of our expert interview indicate that with our adapted version participants require fewer executions of the analysis. Conclusion: Our results indicate that error chains occur frequently in real-world projects, and ignoring them can lead to imprecise evaluation results. The runtime benchmark indicates that our tool is a feasible and efficient solution for detecting error chains in real-world projects. Further, our results gave a hint that the usability of static analyses may benefit from supporting error chains.

cs.SE↗

arXiv2024

Amplifying Exploration in Monte-Carlo Tree Search by Focusing on the Unknown

Monte-Carlo tree search (MCTS) is an effective anytime algorithm with a vast amount of applications. It strategically allocates computational resources to focus on promising segments of the search tree, making it a very attractive search algorithm in large search spaces. However, it often expends its limited resources on reevaluating previously explored regions when they remain the most promising path. Our proposed methodology, denoted as AmEx-MCTS, solves this problem by introducing a novel MCTS formulation. Central to AmEx-MCTS is the decoupling of value updates, visit count updates, and the selected path during the tree search, thereby enabling the exclusion of already explored subtrees or leaves. This segregation preserves the utility of visit counts for both exploration-exploitation balancing and quality metrics within MCTS. The resultant augmentation facilitates in a considerably broader search using identical computational resources, preserving the essential characteristics of MCTS. The expanded coverage not only yields more precise estimations but also proves instrumental in larger and more complex problems. Our empirical evaluation demonstrates the superior performance of AmEx-MCTS, surpassing classical MCTS and related approaches by a substantial margin.

cs.AI↗

arXiv2024

Towards Trustworthy AI Software Development Assistance

It is expected that in the near future, AI software development assistants will play an important role in the software industry. However, current software development assistants tend to be unreliable, often producing incorrect, unsafe, or low-quality code. We seek to resolve these issues by introducing a holistic architecture for constructing, training, and using trustworthy AI software development assistants. In the center of the architecture, there is a foundational LLM trained on datasets representative of real-world coding scenarios and complex software architectures, and fine-tuned on code quality criteria beyond correctness. The LLM will make use of graph-based code representations for advanced semantic comprehension. We envision a knowledge graph integrated into the system to provide up-to-date background knowledge and to enable the assistant to provide appropriate explanations. Finally, a modular framework for constrained decoding will ensure that certain guarantees (e.g., for correctness and security) hold for the generated code.

cs.SE↗

arXiv2023

LoRe: A Programming Model for Verifiably Safe Local-First Software

Local-first software manages and processes private data locally while still enabling collaboration between multiple parties connected via partially unreliable networks. Such software typically involves interactions with users and the execution environment (the outside world). The unpredictability of such interactions paired with their decentralized nature make reasoning about the correctness of local-first software a challenging endeavor. Yet, existing solutions to develop local-first software do not provide support for automated safety guarantees and instead expect developers to reason about concurrent interactions in an environment with unreliable network conditions. We propose LoRe, a programming model and compiler that automatically verifies developer-supplied safety properties for local-first applications. LoRe combines the declarative data flow of reactive programming with static analysis and verification techniques to precisely determine concurrent interactions that violate safety invariants and to selectively employ strong consistency through coordination where required. We propose a formalized proof principle and demonstrate how to automate the process in a prototype implementation that outputs verified executable code. Our evaluation shows that LoRe simplifies the development of safe local-first software when compared to state-of-the-art approaches and that verification times are acceptable.

cs.PL↗

arXiv2023

Using Rewrite Strategies for Efficient Functional Automatic Differentiation

Automatic Differentiation (AD) has become a dominant technique in ML. AD frameworks have first been implemented for imperative languages using tapes. Meanwhile, functional implementations of AD have been developed, often based on dual numbers, which are close to the formal specification of differentiation and hence easier to prove correct. But these papers have focussed on correctness not efficiency. Recently, it was shown how an approach using dual numbers could be made efficient through the right optimizations. Optimizations are highly dependent on order, as one optimization can enable another. It can therefore be useful to have fine-grained control over the scheduling of optimizations. One method expresses compiler optimizations as rewrite rules, whose application can be combined and controlled using strategy languages. Previous work describes the use of term rewriting and strategies to generate high-performance code in a compiler for a functional language. In this work, we implement dual numbers AD in a functional array programming language using rewrite rules and strategy combinators for optimization. We aim to combine the elegance of differentiation using dual numbers with a succinct expression of the optimization schedule using a strategy language. We give preliminary evidence suggesting the viability of the approach on a micro-benchmark.

cs.PL↗

arXiv2023

UNGOML: Automated Classification of unsafe Usages in Go

The Go programming language offers strong protection from memory corruption. As an escape hatch of these protections, it provides the unsafe package. Previous studies identified that this unsafe package is frequently used in real-world code for several purposes, e.g., serialization or casting types. Due to the variety of these reasons, it may be possible to refactor specific usages to avoid potential vulnerabilities. However, the classification of unsafe usages is challenging and requires the context of the call and the program's structure. In this paper, we present the first automated classifier for unsafe usages in Go, UNGOML, to identify what is done with the unsafe package and why it is used. For UNGOML, we built four custom deep learning classifiers trained on a manually labeled data set. We represent Go code as enriched control-flow graphs (CFGs) and solve the label prediction task with one single-vertex and three context-aware classifiers. All three context-aware classifiers achieve a top-1 accuracy of more than 86% for both dimensions, WHAT and WHY. Furthermore, in a set-valued conformal prediction setting, we achieve accuracies of more than 93% with mean label set sizes of 2 for both dimensions. Thus, UNGOML can be used to efficiently filter unsafe usages for use cases such as refactoring or a security audit. UNGOML: https://github.com/stg-tud/ungoml Artifact: https://dx.doi.org/10.6084/m9.figshare.22293052

cs.SE↗

arXiv2023

Towards Code Generation from BDD Test Case Specifications: A Vision

Automatic code generation has recently attracted large attention and is becoming more significant to the software development process. Solutions based on Machine Learning and Artificial Intelligence are being used to increase human and software efficiency in potent and innovative ways. In this paper, we aim to leverage these developments and introduce a novel approach to generating frontend component code for the popular Angular framework. We propose to do this using behavior-driven development test specifications as input to a transformer-based machine learning model. Our approach aims to drastically reduce the development time needed for web applications while potentially increasing software quality and introducing new research ideas toward automatic code generation.

cs.SE↗

arXiv2023

Prisma: A Tierless Language for Enforcing Contract-Client Protocols in Decentralized Applications (Extended Version)

Decentralized applications (dApps) consist of smart contracts that run on blockchains and clients that model collaborating parties. dApps are used to model financial and legal business functionality. Today, contracts and clients are written as separate programs -- in different programming languages -- communicating via send and receive operations. This makes distributed program flow awkward to express and reason about, increasing the potential for mismatches in the client-contract interface, which can be exploited by malicious clients, potentially leading to huge financial losses. In this paper, we present Prisma, a language for tierless decentralized applications, where the contract and its clients are defined in one unit and pairs of send and receive actions that "belong together" are encapsulated into a single direct-style operation, which is executed differently by sending and receiving parties. This enables expressing distributed program flow via standard control flow and renders mismatching communication impossible. We prove formally that our compiler preserves program behavior in presence of an attacker controlling the client code. We systematically compare Prisma with mainstream and advanced programming models for dApps and provide empirical evidence for its expressiveness and performance.

cs.PL↗

arXiv2023

A Direct-Style Effect Notation for Sequential and Parallel Programs

Modeling sequential and parallel composition of effectful computations has been investigated in a variety of languages for a long time. In particular, the popular do-notation provides a lightweight effect embedding for any instance of a monad. Idiom bracket notation, on the other hand, provides an embedding for applicatives. First, while monads force effects to be executed sequentially, ignoring potential for parallelism, applicatives do not support sequential effects. Composing sequential with parallel effects remains an open problem. This is even more of an issue as real programs consist of a combination of both sequential and parallel segments. Second, common notations do not support invoking effects in direct-style, instead forcing a rigid structure upon the code. In this paper, we propose a mixed applicative/monadic notation that retains parallelism where possible, but allows sequentiality where necessary. We leverage a direct-style notation where sequentiality or parallelism is derived from the structure of the code. We provide a mechanisation of our effectful language in Coq and prove that our compilation approach retains the parallelism of the source program.

cs.PL↗

arXiv2023

To Fix or Not to Fix: A Critical Study of Crypto-misuses in the Wild

Recent studies have revealed that 87 % to 96 % of the Android apps using cryptographic APIs have a misuse which may cause security vulnerabilities. As previous studies did not conduct a qualitative examination of the validity and severity of the findings, our objective was to understand the findings in more depth. We analyzed a set of 936 open-source Java applications for cryptographic misuses. Our study reveals that 88.10 % of the analyzed applications fail to use cryptographic APIs securely. Through our manual analysis of a random sample, we gained new insights into effective false positives. For example, every fourth misuse of the frequently misused JCA class MessageDigest is an effective false positive due to its occurrence in a non-security context. As we wanted to gain deeper insights into the security implications of these misuses, we created an extensive vulnerability model for cryptographic API misuses. Our model includes previously undiscussed attacks in the context of cryptographic APIs such as DoS attacks. This model reveals that nearly half of the misuses are of high severity, e.g., hard-coded credentials and potential Man-in-the-Middle attacks.

cs.CR↗

arXiv2022

Dependently-Typed Data Plane Programming

Programming languages like P4 enable specifying the behavior of network data planes in software. However, with increasingly powerful and complex applications running in the network, the risk of faults also increases. Hence, there is growing recognition of the need for methods and tools to statically verify the correctness of P4 code, especially as the language lacks basic safety guarantees. Type systems are a lightweight and compositional way to establish program properties, but there is a significant gap between the kinds of properties that can be proved using simple type systems (e.g., SafeP4) and those that can be obtained using full-blown verification tools (e.g., p4v). In this paper, we close this gap by developing $Π$4, a dependently-typed version of P4 based on decidable refinements. We motivate the design of $Π$4, prove the soundness of its type system, develop an SMT-based implementation, and present case studies that illustrate its applicability to a variety of data plane programs.

cs.PL↗