Semantic Code Clone Research Articles

Semantic Code Clone Detection Based on Community Detection

Semantic code clone detection is to find code snippets that are structurally or syntactically different, but semantically identical. It plays an important role in software reuse, code compression. Many existing studies have achieved good performance in non-semantic clone, but semantic clone is still a challenging task. Recently, several works have used tree or graph, such as Abstract Syntax Tree (AST), Control Flow Graph (CFG) or Program Dependency Graph (PDG) to extract semantic information from source codes. In order to reduce the complexity of tree and graph, some studies transform them into node sequences. However, this transformation will lose some semantic information. To address this issue, we propose a novel high-performance method that utilizes community detection to extract features of AST while preserving its semantic information. First, based on the AST of source code, we exploit community detection to split AST into different subtrees to extract the underlying semantics information of different code blocks, and use centrality analysis to quantify the semantic information as the weight of AST nodes. Then, the AST is converted into a sequence of tokens with weights, and a Siamese neural network model is used to detect the similarity of token sequences for semantic code clone detection. Finally, to evaluate our approach, we conduct experiments on two standard benchmark datasets, Google Code Jam (GCJ) and BigCloneBench (BCB). Experimental results show that our model outperforms the eight publicly available state-of-the-art methods in detecting code clones. It is five times faster than the tree-based method (ASTNN) in terms of time complexity.

Graph-based code semantics learning for efficient semantic code clone detection

Graph-based code semantics learning for efficient semantic code clone detection

Modeling Functional Similarity in Source Code With Graph-Based Siamese Networks

Code clones are duplicate code fragments that share (nearly) similar syntax or semantics. Code clone detection plays an important role in software maintenance, code refactoring, and reuse. A substantial amount of research has been conducted in the past to detect clones. A majority of these approaches use lexical and syntactic information to detect clones. However, only a few of them target semantic clones. Recently, motivated by the success of deep learning models in other fields, including natural language processing and computer vision, researchers have attempted to adopt deep learning techniques to detect code clones. These approaches use lexical information (tokens) and(or) syntactic structures like abstract syntax trees (ASTs) to detect code clones. However, they do not make sufficient use of the available structural and semantic information, hence limiting their capabilities. This paper addresses the problem of semantic code clone detection using program dependency graphs and geometric neural networks, leveraging the structured syntactic and semantic information. We have developed a prototype tool <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Holmes</small> , based on our novel approach and empirically evaluated it on popular code clone benchmarks. Our results show that <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Holmes</small> performs considerably better than the other state-of-the-art tool, TBCCD. We also assessed <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Holmes</small> on unseen projects and performed cross dataset experiments to evaluate the generalizability of <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Holmes</small> . Our results affirm that <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Holmes</small> outperforms TBCCD since most of the pairs that <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Holmes</small> detected were either undetected or suboptimally reported by TBCCD.

Assessing and Improving an Evaluation Dataset for Detecting Semantic Code Clones via Deep Learning

In recent years, applying deep learning to detect semantic code clones has received substantial attention from the research community. Accordingly, various evaluation benchmark datasets, with the most popular one as BigCloneBench, are constructed and selected as benchmarks to assess and compare different deep learning models for detecting semantic clones. However, there is no study to investigate whether an evaluation benchmark dataset such as BigCloneBench is properly used to evaluate models for detecting semantic code clones. In this article, we present an experimental study to show that BigCloneBench typically includes semantic clone pairs that use the same identifier names, which however are not used in non-semantic-clone pairs. Subsequently, we propose an undesirable-by-design Linear-Model that considers only which identifiers appear in a code fragment; this model can achieve high effectiveness for detecting semantic clones when evaluated on BigCloneBench, even comparable to state-of-the-art deep learning models recently proposed for detecting semantic clones. To alleviate these issues, we abstract a subset of the identifier names (including type, variable, and method names) in BigCloneBench to result in AbsBigCloneBench and use AbsBigCloneBench to better assess the effectiveness of deep learning models on the task of detecting semantic clones.

Clone-Seeker: Effective Code Clone Search Using Annotations

Source code search plays an important role in software development, e.g. for exploratory development or opportunistic reuse of existing code from a code base. Often, exploration of different implementations with the same functionality is needed for tasks like automated software transplantation, software diversification, and software repair. Code clones, which are syntactically or semantically similar code fragments, are perfect candidates for such tasks. Searching for code clones involves a given search query to retrieve the relevant code fragments. We propose a novel approach called Clone-Seeker that focuses on utilizing clone class features in retrieving code clones. For this purpose, we generate metadata for each code clone in the form of a natural language document. The metadata includes a pre-processed list of identifiers from the code clones augmented with a list of keywords indicating the semantics of the code clone. This keyword list can be extracted from a manually annotated general description of the clone class, or automatically generated from the source code of the entire clone class. This approach helps developers to perform code clone search based on a search query written either as source code terms, or as natural language. With various experiments, we show that (1) Clone-Seeker is effective in finding clones from BigCloneBench dataset by applying code queries and natural language queries; 2) Clone-Seeker has a higher recall when searching for semantic code clones (i.e., Type-4) in BigCloneBench than the state-of-the-art; 3) Clone-Seeker is a generalized technique as it is effective in finding clones in Project CodeNet dataset by applying code queries and natural language queries. 4) Clone-Seeker with manual annotation outperforms other variants in finding clones on the basis of natural language queries

What You See is What it Means! Semantic Representation Learning of Code based on Visualization and Transfer Learning

Recent successes in training word embeddings for Natural Language Processing ( NLP ) tasks have encouraged a wave of research on representation learning for source code, which builds on similar NLP methods. The overall objective is then to produce code embeddings that capture the maximum of program semantics. State-of-the-art approaches invariably rely on a syntactic representation (i.e., raw lexical tokens, abstract syntax trees, or intermediate representation tokens) to generate embeddings, which are criticized in the literature as non-robust or non-generalizable. In this work, we investigate a novel embedding approach based on the intuition that source code has visual patterns of semantics. We further use these patterns to address the outstanding challenge of identifying semantic code clones. We propose the WySiWiM ( ‘ ‘What You See Is What It Means ” ) approach where visual representations of source code are fed into powerful pre-trained image classification neural networks from the field of computer vision to benefit from the practical advantages of transfer learning. We evaluate the proposed embedding approach on the task of vulnerable code prediction in source code and on two variations of the task of semantic code clone identification: code clone detection (a binary classification problem), and code classification (a multi-classification problem). We show with experiments on the BigCloneBench (Java), Open Judge (C) that although simple, our WySiWiM approach performs as effectively as state-of-the-art approaches such as ASTNN or TBCNN. We also showed with data from NVD and SARD that WySiWiM representation can be used to learn a vulnerable code detector with reasonable performance (accuracy ∼90%). We further explore the influence of different steps in our approach, such as the choice of visual representations or the classification algorithm, to eventually discuss the promises and limitations of this research direction.

Clone detection through srcClone: A program slicing based approach

Software clone detection is an often used approach to understand and maintain software systems. One category of clones that is challenging to detect but very useful is semantic clones, which are similar in semantics but differ in syntax significantly. Semantic clone detectors have trouble scaling to larger systems and sometimes struggle with recall and precision. To address this, we developed a slice-based scalable approach that detects both syntactic and semantic code clones, srcClone. srcClone ascertains code segment similarity by assessing the similarity of their corresponding program slices. We employ a lightweight, publicly-available, scalable program slicer within our clone detection approach. Using dependency analysis to detect and assess cloned components, we discover insights about code components that can be affected by a clone pair or set. These elements are critical in impact analysis. It can also be used by program analysts to run on non-compilable and incomplete source code, which serves comprehension and maintenance tasks very well. We first evaluate srcClone by comparing it to six state-of-the-art tools and two additional semantic clone detectors in performance, recall, and precision. We use the BigCloneBench real clones benchmark to facilitate comparison. We use an additional 16 established benchmark scenarios to perform a qualitative comparison between srcClone and 44 clone detection approaches in their capabilities to detect these scenarios. To further measure scalability, we evaluate srcClone on 191 versions of Linux kernel, containing approximately 87 MLOC. In our evaluations, we illustrate our approach is both relatively scalable and accurate. While its precision is slightly less than some other techniques, it makes up for it in higher recall including semantic clones unable to be found by any existing techniques. We contend our approach is an important advancement in software cloning that it is both demonstrably scalable and can detect more types of clones than existing work, thus providing developers greater information into their software.

A detection framework for semantic code clones and obfuscated code

Code obfuscation is a staple tool in malware creation where code fragments are altered substantially to make them appear different from the original, while keeping the semantics unaffected. A majority of the obfuscated code detection methods use program structure as a signature for detection of unknown codes. They usually ignore the most important feature, which is the semantics of the code, to match two code fragments or programs for obfuscation. Obfuscated code detection is a special case of the semantic code clone detection task. We propose a detection framework for detecting both code obfuscation and clone using machine learning. We use features extracted from Java bytecode dependency graphs (BDG), program dependency graphs (PDG) and abstract syntax trees (AST). BDGs and PDGs are two representations of the semantics or meaning of a Java program. ASTs capture the structural aspects of a program. We use several publicly available code clone and obfuscated code datasets to validate the effectiveness of our framework. We use different assessment parameters to evaluate the detection quality of our proposed model. Experimental results are excellent when compared with contemporary obfuscated code and code clone detectors. Interestingly, we achieve 100% success in detecting obfuscated code based on recall, precision, and F1-Score. When we compare our method with other methods for all of obfuscations types, viz, contraction, expansion, loop transformation and renaming, our model appears to be the winner. In case of clone detection our model achieve very high detection accuracy in comparison to other similar detectors.

Semantic code clone detection for Internet of Things applications using reaching definition and liveness analysis

Knowledge extraction from existing software resources for maintenance, re-engineering and bug removal through code clone detection is an integral part of most of the internet-enabled devices. Similar code fragments which are live at different locations are called code clones. These Internet-enabled devices are used for knowledge sharing and data extraction to execute various applications related to code clone detection. However, most of the existing semantic code clone detection techniques are unable to provide heuristic solution for problems such as statement reordering, inversion of control predicates and insertion of irrelevant statements which may cause a performance bottleneck in this environment. To address these issues, we propose a novel approach that finds semantic code clones in a program or procedure using data flow analysis on the basis of reaching definition and liveness analysis. The algorithm based on reaching definition and liveness analysis is designed to find similar code fragments which are structurally divergent, but semantically equivalent. The results obtained demonstrate that the proposed approach using reaching definition and liveness analysis is effective in detection of semantic code clones for various applications running on the Internet-enabled devices. We have found 5831 semantically equivalent clone pairs on subject systems taken from DeCapo benchmark after elimination of 29,029 dead codes/statements having 2,16,579 line of code (LOC).