Network Flow Attack For Split Manufacturing

By: Yujie Wang (ICT, Chinese Academy of Sciences), Pu Chen (Alexa Shopping, Amazon Web Services), Jiang Hu (Texas A&M University), Guofeng Li (Nankai University), and JV Rajendran (Texas A&M University)

Stage: Physical Design and Manufacturing

Summary

The network flow attack for split manufacturing is an attack for recovering the missing connections in the partial design of a circuit. It is written in Python. Given a DEF file for a circuit, the layer that the circuit is split at, and the appropriate library files, the tool returns a verilog file for the circuit after inferring the missing connections. The tool uses five hints and formulates the problem of finding the missing connections as a network flow problem. The five hints are (1) physical proximity, (2) acyclic combination, (3) capacitance constraints (4) direction of dangling wire, and (5) timing constraints. The tool also provides flexibility whether or not to use the last three hints independently i.e., one can use any of the 2^3 combinations of the last three hints. The solution of the network flow problem is translated into inference about the missing connections using which an attacker can get the complete design of the circuit.

Contact

JV Rajendran and Vasudev Gohil

Input/Output Interface

• Input: DEF file of the circuit, split layer, attack hints to use
• Output: Attacked verilog file of the circuit with missing connections added

Dependencies

Python 3.6 or above

References

@article{8259507,

title = {The Cat and Mouse in Split Manufacturing},

author = {Yujie Wang and Pu Chen and Jiang Hu and Guofeng Li and Jeyavijayan Rajendran},

url = {https://ieeexplore.ieee.org/document/8259507},

doi = {10.1109/TVLSI.2017.2787754},

issn = {1557-9999},

year  = {2018},

date = {2018-05-01},

journal = {IEEE Transactions on Very Large Scale Integration (VLSI) Systems},

volume = {26},

number = {5},

pages = {805-817},

abstract = {Split manufacturing of integrated circuits eliminates vulnerabilities introduced by an untrusted foundry by manufacturing only a part of the target design at an untrusted high-end foundry and the remaining part at a trusted low-end foundry. Most researchers have focused on attack and defenses for hierarchical designs and/or use a relatively high-end trusted foundry, leading to high cost. We propose an attack and defense for split manufacturing for flattened designs. Our attack uses a network-flow model and outperforms previous attacks. We also develop two defense techniques using placement perturbation-one using physical design information and the other using logical information-while considering overhead. The effectiveness of our techniques is demonstrated on benchmark circuits.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

Acknowledgments

• This work was supported in part by NSF under Grant CCF-1618824 and in part by SRC under Grant 2016- TS-2688 and Grant 2016-TS-2689. The work of Y. Wang was supported by the China Scholarship Council