SAT-Resilient Logic Locking for Hardware Intellectual Property Protection

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 258 papers117 conceptsfrozen 2026-10-04

In the modern semiconductor industry, the high cost of fabrication facilities forces most design houses to outsource the manufacturing of their integrated circuits (ICs) to third-party foundries. This outsourcing creates a significant security risk: an untrusted foundry could steal the intellectual property (IP) of the design, produce counterfeit chips, or insert malicious hardware Trojans.

To counter this, researchers have developed logic locking (or logic encryption). The goal is to hide the circuit's function by incorporating "key" inputs. Without the correct secret key, the circuit produces incorrect outputs, effectively rendering the design useless to an attacker who only possesses the netlist.

The SAT Attack: Pruning the Search Space

For years, the "gold standard" for breaking these defenses has been the SAT attack. As Bicky Shakya and Xiaolin Xu describe in their benchmarking work, a SAT attack is an "oracle-guided" process. The attacker has two things: the locked netlist (from the foundry) and a functional, activated chip (an "oracle") purchased from the open market.

The SAT solver doesn't just guess keys. Instead, it iteratively finds Distinguishing Input Patterns (DIPs)—specific inputs for which two different potential keys produce different outputs. By comparing these outputs against the functional oracle, the solver can rule out massive swaths of the key space simultaneously. This pruning is so efficient that traditional logic locking, which might seem secure due to a large number of key bits, can often be cracked in seconds.

Forcing Exponential Complexity

The core of modern research in this area is finding ways to force the SAT solver back into a brute-force search. If a solver can only rule out one incorrect key per iteration, the attack time becomes exponential relative to the key length, making it computationally infeasible.

Bicky Shakya and Domenic Forte introduced CAS-Lock to achieve this. Unlike traditional tree-based architectures, CAS-Lock uses a "daisy-chained" sequence of key-controlled gates. This specific topology ensures that each incorrect key is invalidated by only one unique input pattern. This creates a mathematical lower bound of security: the solver is forced to evaluate the entire input space, effectively neutralizing the SAT solver's pruning capabilities.

Reconfigurable Fabrics as Black Boxes

A more physical approach to logic locking is redaction, where critical parts of the circuit are replaced with reconfigurable logic, such as an embedded FPGA (eFPGA). To an attacker, these modules appear as "black boxes" with unknown internal routing.

However, Jitendra Bhandari and Benjamin Tan have shown that simply adding an eFPGA is not a "silver bullet." Their research on the ALICE and ARIANNA frameworks reveals a counter-intuitive truth: bitstream length does not equal security.

Instead, security is determined by the complexity of the internal routing network and the logic block density. Attackers can often recover secret bitstreams by "partially unrolling" the combinational loops within the fabric. If the routing is too simple, the SAT solver can resolve the logic even if it only models a small fraction of the circuit's total complexity. To be truly secure, the fabric must reach a "complexity threshold" where the number of possible routing configurations creates a bottleneck for the Boolean solver.

The Scan Chain Leak

Even the most robust logic locking can be undermined by standard "Design-for-Testability" features. Scan chains are internal shift registers used to test chips after manufacturing, but they inadvertently provide a high-bandwidth interface for observing internal states.

Chengkang He and Chip Hong Chang demonstrated that data analytics can identify Finite State Machine (FSM) registers through these scan chains, leaking the chip's control logic. To defend against this, researchers like Seetal Potluri have proposed SeqL+, which scrambles the scan-chain sequence. By modeling the obfuscation as a graph problem, they force an attacker to solve for a Hamiltonian Path in a search space that grows exponentially with the number of flip-flops.

Beyond the Netlist

While much of the field focuses on "netlist-only" attacks, researchers like Mohamed El Massad and Rupesh Raj Karn argue that true security must account for the presence of an oracle.

The next frontier involves integrating structural obfuscation—making the netlist look like a jumbled mess—with functional defenses that prevent an attacker from using an activated chip to "query" the secret key. As AJ Fowler and Yiorgos Makris suggest with their work on TRAP (Transistor-Level Programmable fabrics), moving the defense from the gate level down to the physical transistor level may be the only way to stay ahead of increasingly sophisticated mathematical solvers.

Papers in this version

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 65 papers
  2. Version 1 · 2026-10-04 · 58 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_sat_based_reverse_engineering_c9bd33ee,
  title = {SAT-based reverse engineering},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/sat-based-reverse-engineering/rdr_c9bd33eecd72403cbd29e1dccb18eff2},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_sat_resilient_logic_locking_for_hardware_c9bd33ee_v1,
  title = {SAT-Resilient Logic Locking for Hardware Intellectual Property Protection},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/sat-resilient-logic-locking-for-hardware-intellectual-property-protection/rdr_c9bd33eecd72403cbd29e1dccb18eff2/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}