IKOS
IKOS (Inference Kernel for Open Static Analyzers) is a static analyzer for C/C++ based on the theory of Abstract Interpretation.
Introduction
IKOS started as a C++ library designed to facilitate the development of sound static analyzers based on Abstract Interpretation. Specialization of a static analyzer for an application or family of applications is critical for achieving both precision and scalability. Developing such an analyzer is arduous and requires significant expertise in Abstract Interpretation.
IKOS provides a generic and efficient implementation of state-of-the-art Abstract Interpretation data structures and algorithms, such as control-flow graphs, fixpoint iterators, numerical abstract domains, etc. IKOS is independent of a particular programming language.
IKOS also provides a C and C++ static analyzer based on LLVM. It implements scalable analyses for detecting and proving the absence of runtime errors in C and C++ programs.
License
IKOS has been released under the NASA Open Source Agreement version 1.3, see LICENSE.pdf
Contact
Release notes
See RELEASE_NOTES.md
Troubleshooting
Installation
Dependencies
To build and run the analyzer, you will need the following dependencies:
- A C++ compiler that supports C++14 (gcc >= 4.9.2 or clang >= 3.4)
- CMake >= 3.4.3
- GMP >= 4.3.1
- Boost >= 1.55
- Python 2 >= 2.7.3 or Python 3 >= 3.3
- SQLite >= 3.6.20
- TBB >= 2
- LLVM and Clang 9.0.x
- (Optional) APRON >= 0.9.10
- (Optional) Pygments
Most of them can be installed using your package manager.
Installation instructions for Arch Linux, CentOS, Debian, Fedora, Mac OS X, Red Hat, Ubuntu and Windows are available in the doc/install directory. These instructions assume you have sudo or root access. If you don't, please follow the instructions in doc/install/ROOTLESS.md.
Note: If you build LLVM from source, you need to enable run-time type information (RTTI).
Once you have all the required dependencies, move to the next section.
Build and Install
Now that you have all the dependencies on your system, you can build and install IKOS.
As you open the IKOS distribution, you shall see the following directory structure:
.
├── CMakeLists.txt
├── LICENSE.pdf
├── README.md
├── RELEASE_NOTES.md
├── TROUBLESHOOTING.md
├── analyzer
├── ar
├── cmake
├── core
├── doc
├── frontend
├── script
└── test
IKOS uses the CMake build system. You will need to specify an installation directory that will contain all the binaries, libraries and headers after installation. If you do not specify this directory, CMake will install everything under install in the root directory of the distribution. In the following steps, we will install IKOS under /path/to/ikos-install-directory.
Here are the steps to build and install IKOS:
$ mkdir build
$ cd build
$ cmake -DCMAKE_INSTALL_PREFIX=/path/to/ikos-install-directory ..
$ make
$ make install
Then, add IKOS in your PATH (consider adding this in your .bashrc):
$ PATH="/path/to/ikos-install-directory/bin:$PATH"
Tests
To build and run the tests, simply type:
$ make check
How to run IKOS
Suppose we want to analyze the following C program in a file, called loop.c:
1: #include <stdio.h>
2: int a[10];
3: int main(int argc, char *argv[]) {
4: size_t i = 0;
5: for (;i < 10; i++) {
6: a[i] = i;
7: }
8: a[i] = i;
9: printf("%i", a[i]);
10: }
To analyze this program with IKOS, simply run:
$ ikos loop.c
You shall see the following output. IKOS reports two occurrences of buffer overflow at line 8 and 9.
[*] Compiling loop.c
[*] Running ikos preprocessor
[*] Running ikos analyzer
[*] Translating LLVM bitcode to AR
[*] Running liveness analysis
[*] Running widening hint analysis
[*] Running interprocedural value analysis
[*] Analyzing entry point 'main'
[*] Checking properties for entry point 'main'
# Time stats:
clang : 0.037 sec
ikos-analyzer: 0.023 sec
ikos-pp : 0.007 sec
# Summary:
Total number of checks : 7
Total number of unreachable checks : 0
Total number of safe checks : 5
Total number of definite unsafe checks: 2
Total number of warnings : 0
The program is definitely UNSAFE
# Results
loop.c: In function 'main':
loop.c:8:10: error: buffer overflow, trying to access index 10 of global variable 'a' of 10 elements
a[i] = i;
^
loop.c: In function 'main':
loop.c:9:18: error: buffer overflow, trying to access index 10 of global variable 'a' of 10 elements
printf("%i", a[i]);
^
The ikos command takes a source file (.c, .cpp) or a LLVM bitcode file (.bc) as input, analyzes it to find runtime errors (also called undefined behaviors), creates a result database output.db in the current working directory and prints a report.
In the report, each line has one of the following status:
- safe: the statement is proven safe;
- error: the statement always results into an error (or is unreachable);
- unreachable: the statement is never executed;
- warning may mean three things:
- the statement results into an error for some executions, or
- the static analyzer did not have enough information to conclude (check dependent on an external input, for instance), or
- the static analyzer was not powerful enough to prove the absence of errors;
By default, ikos shows warnings and errors directly in your terminal, like a compiler would do.
If the analysis report is too big, you shall use:
ikos-report output.dbto examine the report in your terminalikos-view output.dbto examine the report in a web interface
Further information:
- Analyze a whole project with ikos-scan
- Examine a report with ikos-view
- Analysis Options
- Report Options
- APRON Support
- Analysis Assumptions
- Analyze an embedded software requiring a cross-compiler
- Model library functions to reduce warnings
Contributors
See CONTRIBUTORS.md
Publications
-
Sung Kook Kim, Arnaud J. Venet, Aditya V. Thakur. Deterministic Parallel Fixpoint Computation. In Principles of Programming Languages (POPL 2020), New Orleans, Louisiana (PDF).
-
Guillaume Brat, Jorge Navas, Nija Shi and Arnaud Venet. IKOS: a Framework for Static Analysis based on Abstract Interpretation. In Proceedings of the International Conference on Software Engineering and Formal Methods (SEFM 2014), Grenoble, France (PDF).
-
Arnaud Venet. The Gauge Domain: Scalable Analysis of Linear Inequality Invariants. In Proceedings of Computer Aided Verification (CAV 2012), Berkeley, California, USA 2012. Lecture Notes in Computer Science, pages 139-154, volume 7358, Springer 2012 (PDF).
Coding Standards
Overview of the source code
See doc/OVERVIEW.md
555 Jan 8, 2023
13.8k Jan 4, 2023
1.2k Jan 7, 2023
1.3k Dec 18, 2022
18 Aug 30, 2022
21 Dec 21, 2022
44 Sep 3, 2022
4.5k Dec 30, 2022
1.8k Dec 28, 2022
160 Dec 26, 2022
6.6k Jan 8, 2023
5 Dec 15, 2021
268 Jan 9, 2023
1 Nov 13, 2021
4.4k Jan 7, 2023
119 Dec 5, 2022
1 Oct 16, 2021
14 Sep 19, 2022