Just-in-time software vulnerability detection: Are we there yet?

Lomio, Francesco; Iannone, Emanuele; Lucia, Andrea De; Palomba, Fabio; Lenarduzzi, Valentina

doi:10.1016/j.jss.2022.111283

Cited by 39 publications

(15 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…2) Generalization to Diverse Vulnerabilities: Current vulnerability prediction methods often focus on specific types of vulnerabilities or employ a single generic model to handle multiple vulnerabilities. However, this approach may limit predictive performance and generalization when dealing with diverse and complex vulnerabilities [30]. Our MTLPT method focuses on the five most frequent vulnerability types in the real world, enhancing the model's understanding of both differences and commonalities among vulnerabilities through multi-task learning.…”

Section: Discussionmentioning

confidence: 99%

“…The basic idea of ensemble learning is: a group of "experts" (i.e., different models) collectively make decisions that are usually better than any single expert's decision. Ensemble learning methods are mainly divided into two categories: Bagging [27] (Bootstrap Aggregating) such as Random Forest [28] (RF), Boosting [29] such as: AdaBoost [30], they all improve the overall accuracy by combining the prediction results of multiple simple models. For example, using multiple decision trees to vote to determine whether a code snippet may contain vulnerabilities [29].…”

Section: Ensemble Learning Methodsmentioning

confidence: 99%

“…Vulnerabilities are data with complex features involving multiple dimensions of source code, including syntactic structure, semantic information, control flow, data flow, and developers' coding habits. They are characterized by semantic complexity, high dimensionality, data sparsity, and imbalance [30]. Vulnerabilities in code also have complex dependencies; they may not only depend on local code snippets but also involve complex dependencies across modules and libraries [33].…”

Section: Model Architecturementioning

confidence: 99%

See 2 more Smart Citations

Elevating Security Analysis: The MTLPT Framework for Enhanced Vulnerability Prediction

HUI,

LIU,

CHEN

et al. 2024

Preprint

View full text Add to dashboard Cite

In the current field of vulnerability prediction, accurate forecasting and identification of potential vulnerabilities in software are crucial, especially when dealing with real-world vulnerability data. The challenges posed by data imbalance and complex dependency relationships often make prediction tasks exceptionally difficult. Traditional single-task learning methods and ensemble learning methods typically perform poorly when handling highly imbalanced datasets. These methods often overlook minority categories, which frequently contain the most critical vulnerabilities, and fail to fully learn vulnerability features. To address these issues, we propose a novel multi-task learning model called MTLPT, aimed at enhancing the accuracy and efficiency of vulnerability prediction through a multi-task learning framework. The MTLPT model combines custom lightweight Transformer blocks and position encoding layers to effectively capture long-range dependencies and complex contextual information from source code. With this structural design, MTLPT can simultaneously handle various vulnerability prediction tasks, thereby learning the latent relationships between different vulnerability types and improving the model’s sensitivity to rare but severe vulnerabilities. Additionally, the MTLPT model introduces a loss function based on dynamic weights, which dynamically adjusts loss weights based on the prediction difficulty of different tasks, effectively mitigating the challenges posed by imbalanced data. We conducted comparative experiments on a subset of highly imbalanced real-world vulnerability dataset. The experimental results demonstrate that, compared to existing single-task learning and ensemble learning methods, MTLPT exhibits significant advantages across multiple key performance metrics, particularly in identifying minority class vulnerabilities with higher sensitivity and accuracy. This performance enhancement validates the effectiveness of our proposed multi-task learning framework in handling complex and imbalanced vulnerability data, highlighting the importance of the MTLPT framework, custom lightweight Transformer blocks, position encoding layers, and the dynamic weight loss function in practical applications. Furthermore, we conducted a series of ablation experiments to thoroughly evaluate the contributions of individual components within the MTLPT model, confirming the role of custom lightweight Transformer blocks and position encoding layers in enhancing the model’s ability to learn complex code structures and behavioral patterns, while also demonstrating the critical role of the dynamic weight loss function in optimizing the training process of the multi-task learning model.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Ensemble Learning Methodsmentioning

confidence: 99%

Section: Model Architecturementioning

confidence: 99%

See 1 more Smart Citation

Elevating Security Analysis: The MTLPT Framework for Enhanced Vulnerability Prediction

HUI,

LIU,

CHEN

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…In some cases metrics like the code's entropy, number of characters, number of conditional sentences, among others, are applied [20], [28]. Other works also play with cyclomatic complexity [9], [20], use dependency and control flow graphs [10], [22], [24], [25], [27], [30]- [34], even work at token level [23], use vectors to represent assorted information [8], [19], [21], [26] or directly apply the code [29]. Once features are extracted, machine learning algorithms are used in all cases, being neural networks the most common one, specially MLP, though some works also use DNN, like [29].…”

Section: A Vulnerability Detectionmentioning

confidence: 99%

Vulnerability detection under poisoning attacks through code and token features

González-Manzano,

Garcia-Alfaro

2024

Preprint

View full text Add to dashboard Cite

The complexity of implementations and the interconnection of assorted systems and devices facilitates the emergence of vulnerabilities. Detection systems are developed to fight against this security issue, being the use of Artificial Intelligence (AI) a common practice. However, the use of AI is not without its problems, specially those affecting the training phase. This paper tackles this issue following a two-fold approach. First, an AI-based vulnerability detection system based on code and token metrics, dubbed VulCoT, is developed. It reaches state-of-the-art performance while being suitable for C#, C/C++ and PHP. Second, the impact of poisoning attacks on VulCoT is analysed. Results show that VulCoT is specially affected beyond 20% of false data. Remarkably, detecting some of the most frequent Common Weakness Enumeration is altered even with lower poison rates. Overall, KNN and SVM are more appropriate for system protection in C# and C/C++, while MLP in PHP. Indeed, PHP is the language which is less affected by attacks, while C# and C/C++ present comparable results.

show abstract

“…S OFTWARE exploitation and protection mechanisms are considered an arms race between attackers and defenders. The prevalence of software vulnerabilities and weak memory protection practices allow attackers to corrupt the memory space of the vulnerable applications to run a malicious arbitrary code [1]. Memory safety vulnerabilities are made possible due to the inherent characteristics of modernday computing architecture.…”

Section: Introductionmentioning

confidence: 99%

NG-MVEE: A New Proposed Hybrid Technique for Enhanced Mitigation of Code Re-Use Attack

et al. 2023

View full text Add to dashboard Cite

Code-Reuse Attacks (CRAs) are solid mechanisms to bypass advanced software and hardware defenses because they use the software's own code and they are very hard to be detected without significant overhead. Numerous methods have been proposed to protect against memory-based attacks that result from reusing parts of the attacked binary code. In this paper, two problems were tackled. the first problem is the lack of a categorized survey, analysis, and evaluation of the different CRAs proposed in the literature. The second problem is the inherent vulnerability that exists in protection techniques that are based on Multi-Variant Execution Environment (MVEE) since they are using shared Linux libraries with gadget-prone codes. In the paper a novel framework of CRA mitigation is introduced; fusing the two different prominent techniques of control flow integrity and multi-variant execution environment. The novel mitigation technique, named Next Generation MVEE (NG-MVEE), was built upon an existing generic CRA detection system (GHUMVEE) and complemented with a different CRA detection technique (G-Free) in order to provide comprehensive protection against code-reuse attacks. The outcome of the hybrid system is an optimized hybrid version of an MVEE technique, with minimal performance overhead increase due to the added protection layer of the G-Free technique. A median of 7% performance overhead resulted from the proposed protection system.

show abstract

Just-in-time software vulnerability detection: Are we there yet?

Cited by 39 publications

References 46 publications

Elevating Security Analysis: The MTLPT Framework for Enhanced Vulnerability Prediction

Elevating Security Analysis: The MTLPT Framework for Enhanced Vulnerability Prediction

Vulnerability detection under poisoning attacks through code and token features

NG-MVEE: A New Proposed Hybrid Technique for Enhanced Mitigation of Code Re-Use Attack

Contact Info

Product

Resources

About