Automatically patching errors in deployed software

By Jeff H. Perkins, Sunghun Kim, Samuel Larsen, Saman P. Amarasinghe, Jonathan Bachrach, Michael Carbin, Carlos Pacheco, F. Sherwood, Stelios Sidiroglou, Greg Sullivan and 4 more. Symposium on Operating Systems Principles, 2009

432 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Networks and Communications, Signal Processing

View Paper: https://www.semanticscholar.org/paper/74fda58710e2569a4a5ea4a606d8083daab0248f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Networks and Communications

In Signal Processing

Citations in the corpus

Nopol: Automatic Repair of Conditional Statement Bugs in Java Programs (2018, methodology): “Different kinds of techniques are proposed for automatic repair, including patch generation [27], [41] and dynamic program state recovery [42], [7].”
History Driven Program Repair (2016, background): “propose ClearView, targeting security errors in binary programs [40].”
Automatic patch generation by learning correct code (2016, background): “ClearView: ClearView is a generate-and-validate system that observes normal executions to learn invariants that characterize safe behavior [25].”
The ManyBugs and IntroClass Benchmarks for Automated Repair of C Programs (2015, methodology): “Similarly Jolt [68] and ClearView [15] are language independent, address defect types that are included in either or both of the defect sets, and require indicative or runtime workloads.”
An analysis of patch plausibility and correctness for generate-and-validate patch generation systems (2015, methodology): “ClearView is a generate-and-validate system that observes normal executions to learn invariants that characterize safe behavior [30].”
The plastic surgery hypothesis (2014, neutral): “In 2009, the advent of GenProg [41] and Clearview [31] demonstrated automated bug repair.”
DirectFix: Looking for Simple Program Repairs (2015, background): “A number of them repair specific defect types [24]–[30], while DirectFix is designed to be a general purpose program repair tool.”
Towards secure and dependable software-defined networks (2013, background): “Solutions as those proposed in [30] can help in achieving this goal.”
SemFix: Program repair via semantic analysis (2013, methodology): “ClearView [24] follows a similar scheme but works on deployed binary program when high availability is required.”
Staged program repair with condition synthesis (2015, background): “In the last decade researchers have developed automatic techniques that have been demonstrated to successfully correct targeted but important classes of defects such as out of bounds accesses [33, 30, 34], integer overflow errors [34], null pointer dereferences [11, 23], infinite loops [18], memory ”
Neural Nets Can Learn Function Type Signatures From Binaries (2017, background): “Some examples of the tasks include hot patching [33] and fine-”
A systematic study of automated program repair: Fixing 55 out of 105 bugs for $8 each (2012, methodology): “This evaluation includes two orders of magnitude more source code than AutoFix-E [9] or our previous work [11], two orders of magnitude more test cases than ClearView [8], and two orders of magnitude more defects than AFix [10] (in addition to being strictly larger than each of those four previous p”
Large Language Models for Software Engineering: Survey and Open Problems (2023, background): “However, more recently, research on program synthesis took a different turn: Inspired by Genetic Programming [165], and early results from Automated Program Repair [138], [166], it considered a wider set of transformations in an approach that has come to be known as ‘Genetic Improvement’ [7], [167].”
Angelix: Scalable Multiline Program Patch Synthesis via Symbolic Analysis (2016, background): “Apart from general-purpose repair tools like ours, there are also other repair approaches targeted for speciﬁc types of defects (e.g., buﬀer overﬂow) or speciﬁc application domains (e.g., web applications) [5, 9, 11, 18, 27, 30, 35–37].”
A Large-Scale Study on Repetitiveness, Containment, and Composability of Routines in Open-Source Projects (2016, methodology): “First, the automated program repairing approaches [22, 54] involve searching a large space of programs in the codebase with the assumption that a ﬁx might already occur in the same program or other ones [22].”
GenProg: A Generic Method for Automatic Software Repair (2012, background): “We follow Rinard et al. [14] in defining a repair as a patch consisting of one or more code changes that, when applied to a program, cause it to pass a set of test cases (typically including both tests of required behavior as well as a test case encoding the bug).”
Is the cure worse than the disease? overfitting in automated program repair (2015, methodology): “INTRODUCTION Automated program repair [4,13,17,18,26,28,29,32,36,37,39,39, 43,46,47,50,55,58,60,61] holds great potential to reduce debugging costs and improve software quality.”
An Empirical Study on Learning Bug-Fixing Patches in the Wild via Neural Machine Translation (2018, background): “A complementary set of repair techniques leverage program analysis and program synthesis to repair programs by constructing code with particular properties [35, 40, 51, 52, 60, 63, 77, 92].”
Automated program repair (2019, background): “Other approaches consider alternative oracles, such as potential invariants inferred from dynamic executions [32].”