secure-software-engineering
Sparse Analysis Optimization Of Low Code Intrusion
Can improve speed and reduce memory usage. ( Be able to reduce propgeate edges by 20 times
Now this optimization will cause some test cases to fail and is still being resolved
[ERROR] ImplicitFlowTests>ImplicitFlowTests.userCodeTest2:667->JUnitTests.negativeCheckInfoflow:146 expected:<0> but was:<1>
[ERROR] ImplicitFlowTests>ImplicitFlowTests.implicitFlowTaintWrapperNegativeTest:483->JUnitTests.negativeCheckInfoflow:146 expected:<0> but was:<1>
[ERROR] ImplicitFlowTests>ImplicitFlowTests.userCodeTest1:654->JUnitTests.negativeCheckInfoflow:146 expected:<0> but was:<1>
[ERROR] ImplicitFlowTests>ImplicitFlowTests.classTypeTest:583->JUnitTests.checkInfoflow:124 expected:<1> but was:<0>
[ERROR] ImplicitFlowTests>ImplicitFlowTests.switchTest:85->JUnitTests.checkInfoflow:124 expected:<1> but was:<0>
[ERROR] ImplicitFlowTests>ImplicitFlowTests.arrayIndexTest1:620->JUnitTests.checkInfoflow:124 expected:<1> but was:<0>
[ERROR] ImplicitFlowTests>ImplicitFlowTests.conditionalAliasingTest2:516->JUnitTests.negativeCheckInfoflow:146 expected:<0> but was:<1>
[ERROR] ImplicitFlowTests>ImplicitFlowTests.afterCallNegativeTest:550->JUnitTests.negativeCheckInfoflow:146 expected:<0> but was:<1>
If you can ensure that all test cases still work with your optimization and can show a significant speedup, that would be greatly appreciated. Please provide precise data on your measurements, i.e., on which apps with which configuration you tested.
Concerning the implicit flows, you could propgate them non-sparse. Since these are control-flow dependencies, a lot of locals will be tainted anyway, so I guess you won't see much of a speedup there with a sparse propagation. It would then maybe make sense to distinguish between explicit flows (wich you propagate through a sparse graph) and implicit flows (which you propgate as normal). That would also resolve the correctness issues from the test cases.
