RTFM! Automatic Assumption Discovery and Verification Derivation from Library Document for API Misuse Detection

Lv, Tao; Li, Ruishi; Yang, Yi; Chen, Kai; Liao, Xiaojing; Hu, Peiwei; Xing, Luyi

doi:10.1145/3372297.3423360

Cited by 15 publications

(11 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The overlapping ratio is the number of shared rules divided by the minimum number of rules in the three libraries. For example, 12 of the dtype rules are the same across the libraries, resulting in an overlapping ratio of 31.6%, which is 12 𝑚𝑖𝑛 (41,38,47) , since at most only 𝑚𝑖𝑛 (41,38,47) number of rules can be shared across the libraries.…”

Section: Constraint Extraction Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

“…Analyzing software text to detect bugs: Prior work leverages documents [16,38] and comments [57-59, 76, 78] to detect inconsistency bugs between code and its specifications. Some prior work leverages UML statecharts for test case generation [45] and translates software specifications into assertions [37], verificatino code [38], and oracles [25,41]. Different from these techniques, D2C uses sequential pattern mining to aid the extraction of constraints from API documents to guide input generation for testing DL libraries.…”

Section: Related Workmentioning

confidence: 99%

“…In addition, it cannot extract numerical constraints such as shape and range. Existing techniques that extract constraints from software text extract different types of constraints that are not DL-specific such as exceptions [13,25,47,72], command-line options and file formats [69], lock-and call-relations [38,57], interrupts [58], nullness [38,47,59,78], resources [76], dependency relations for web service parameters [70], and inheritance relations [78]. Although some [13,47,72,78] can extract constraints related to valid ranges, those are only a small portion of DL-specific constraints (Section 4.1).…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

DocTer: Documentation Guided Fuzzing for Testing Deep Learning API Functions

Xie,

Li,

Kim

et al. 2021

Preprint

View full text Add to dashboard Cite

It is integral to test API functions of widely-used deep learning (DL) libraries. The effectiveness of such testing requires DL-specific input constraints of these API functions. Such constraints enable the generation of valid inputs, i.e., inputs that follow these DL-specific constraints, to explore deep to test the core functionality of API functions. Existing fuzzers have no knowledge of such constraints, and existing constraint-extraction techniques are ineffective for extracting DL-specific input constraints.To fill this gap, we design and implement a document-guided fuzzing technique-D2C-for API functions of DL libraries. D2C leverages sequential pattern mining to generate rules for extracting DL-specific constraints from API documents and uses these constraints to guide the fuzzing to generate valid inputs automatically. D2C also generates inputs that violate these constraints to test the input validity checking code. In addition, D2C uses the constraints to generate boundary inputs to detect more bugs.Our evaluation of three popular DL libraries (TensorFlow, Py-Torch, and MXNet) shows that D2C's accuracy in extracting input constraints is 83.3-90.0%. D2C detects 121 bugs, while a baseline fuzzer without input constraints detects only 68 bugs. Most (89) of the 121 bugs are previously unknown, 54 of which have been fixed or confirmed by developers after we report them. In addition, D2C detects 38 inconsistencies within documents, including 28 that are fixed or confirmed after we report them.

show abstract

Section: Constraint Extraction Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

DocTer: Documentation Guided Fuzzing for Testing Deep Learning API Functions

Xie,

Li,

Kim

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Then during JIT compilation, PYE uses the library's pre-calculated condition values to parse the conditions related to these values to generate the final result. Tao et al proposed a technique called Advance (Lv et al, 2020), which is used to verify some inviolable properties in the Java API library, such as null dereferences.…”

Section: Null Dereference Analysismentioning

confidence: 99%

Optimizing demand‐driven null dereference verification via merging branches

Huihui

Zeng

2021

Expert Systems

View full text Add to dashboard Cite

Null dereference is a common type of runtime failure in Java programs, and it is necessary to verify whether a dereference in the program is safe. However, previous works often have redundant path exploration and high false positive rate. In this paper, we propose a merged null dereference verification (MNDV) approach. MNDV employs a backward, path‐sensitive inter‐procedural analysis technique to verify a given dereference as safe or potentially unsafe. It uses a branch merging strategy to remove redundant paths, and a method call's relevance to the null references is checked to determine whether it is necessary to explore the internal codes of the method. We have evaluated the approach in some standard benchmark programs. Compared with some existing approaches, our approach reduces false alarm rate and effectively reduce time and memory consumption.

show abstract

NLP-based Cross-Layer 5G Vulnerabilities Detection via Fuzzing Generated Run-Time Profiling

Wang,

Wang

2023

2023 IEEE 12th International Conference on Cloud Networking (CloudNet)

View full text Add to dashboard Cite

RTFM! Automatic Assumption Discovery and Verification Derivation from Library Document for API Misuse Detection

Cited by 15 publications

References 21 publications

DocTer: Documentation Guided Fuzzing for Testing Deep Learning API Functions

DocTer: Documentation Guided Fuzzing for Testing Deep Learning API Functions

Optimizing demand‐driven null dereference verification via merging branches

NLP-based Cross-Layer 5G Vulnerabilities Detection via Fuzzing Generated Run-Time Profiling

Contact Info

Product

Resources

About