Research on Modularization-based Code Reuse Technology in Software System Development
Sep 25, 2025
About this article
Published Online: Sep 25, 2025
Received: Jan 16, 2025
Accepted: May 06, 2025
DOI: https://doi.org/10.2478/amns-2025-1013
Keywords
© 2025 Yu Hu, published by Sciendo
This work is licensed under the Creative Commons Attribution 4.0 International License.
Figure 1.
Figure 2.
Figure 3.
Figure 4.
Figure 5.
Applications and data sets
| Applications | Loophole | The number of benign gadget chains | The number of malicious gadget chains |
|---|---|---|---|
| Adobe flash 11.2.202.336 | CVE-2014-0502 | 201166 | 164523 |
| Nginx 1.4.0 | CVE-2013-2028 | 125364 | 103745 |
| Proftpd 1.3.0a | CVE-2006-6563 | 91365 | 60132 |
| Firefox 3.5.10 | CVE-2010-1214 | 212635 | 186635 |
Different procedures test results
| Applications | Data set | Training data | FP | FN | Accuracy |
|---|---|---|---|---|---|
| Adobe flash | 368777 | 295039 | 0.42% | 0.54% | 97.1% |
| Nginx | 228855 | 183105 | 0.04% | 0.65% | 98.0% |
| Proftpd | 152176 | 121760 | 0.32% | 0.96% | 98.3% |
| Firefox | 407529 | 326042 | 0.06% | 0.38% | 99.5% |
| Average | - | - | 0.14% | 0.54% | 98.0% |
Code length variation
| Code Size/SPEC | Sw ift Code(KB | PIC Code(KB) | #Compiled M ethods | Average Inc.(B) |
|---|---|---|---|---|
| compress | 353.87 | 376.71 | 459 | 54 |
| jess | 610.78 | 652.41 | 863 | 50 |
| db | 381.67 | 404.97 | 478 | 53 |
| javac | 1050.04 | 1099.67 | 1266 | 43 |
| mtrt | 510.06 | 534.33 | 616 | 41 |
| jack | 684.34 | 734 | 712 | 75 |
| Sum/Avg | 3591.08 | 3802.79 | 4383 | 52 |
Time overhead of SPEC CPU2006 Benchmark
| Benchmark | Original running time(*s) | New running time(*s) | Overhead |
|---|---|---|---|
| perlbench | 246 | 246 | 0 |
| bzip2 | 394 | 394 | 0 |
| gcc | 225 | 225 | 0 |
| mcf | 296 | 296 | 0 |
| gobmk | 413 | 413 | 0 |
| hmmer | 317 | 317 | 0 |
| sjeng | 451 | 451 | 0 |
| libquantum | 280 | 280 | 0 |
| h264ref | 443 | 443 | 0 |
| omnetpp | 335 | 335 | 0 |
| astar | 393 | 393 | 0 |
| xalancbmk | 212 | 212 | 0 |
Comparison of different CFI mechanisms
| Safety level | ISA extension | Compiler change | Runtime overhead | Non-invasive detection | |
|---|---|---|---|---|---|
| CFI | I | N | N | 24% | N |
| (CCS’05) | |||||
| CCFI | I | Y | Y | 3%~6% | N |
| (CCS’15) | |||||
| CodeArmor | III | N | N | 6.7% | N |
| (S&P’17) | |||||
| HAFIX | I | Y | Y | 3.1% | N |
| (DAC’15) | |||||
| HCIC | II | N | N | 0.94% | N |
| (JIOT’18) | |||||
| μCFI | III | N | Y | 13% | N |
| (CCS’18) | |||||
| Ours | III | N | N | 0% | Y |
Comparison of neural networks with SVM and LR
| Method | FP | FN | Accuracy |
|---|---|---|---|
| DNN | 0.05% | 0.6% | 99% |
| LR | 8.3% | 20.3% | 82.1% |
| SVM | 7.7% | 25.1% | 74.6% |