SenTrojan: Revealing Analog Hardware Trojan Attacks on Sensors

Kun Wang, Ruochen Zhou, Wenjun Zhu, Shilin Xiao, Kai Wang, Chen Yan, Xiaoyu Ji, Wenyuan XuarXiv 2026 · arXiv

TL;DR

SenTrojan is the first systematic framework for implanting analog hardware Trojans into sensors during the design or manufacturing process. By embedding malicious logic triggered by specific physical signals (like ultrasound or EMI), an attacker can remotely induce persistent spoofing or permanent denial-of-service while maintaining near-perfect stealth during normal operation.

What Problem Does This Paper Address?

Sensors are the critical eyes and ears of cyber-physical systems (CPS), from autonomous vehicles to industrial controllers. Traditionally, sensor security research has focused on signal injection attacks that exploit unintentional hardware vulnerabilities, such as electromagnetic interference (EMI) susceptibility. However, these flaws are often inconsistent across different hardware versions and environments, making attacks difficult to generalize.

SenTrojan shifts the threat model to intentional supply-chain compromise. The authors ask: What if an adversary deliberately implants a hidden attack surface into the sensor's analog circuitry before it is even deployed? The challenge lies in creating a Trojan that is reliable enough to be triggered by noisy physical signals, yet stealthy enough to pass rigorous factory testing without degrading the sensor's accuracy.

Methods

The authors propose two complementary design paradigms for analog Trojans, illustrated in the framework overview below:

Fig. 3: An illustration of the SenTrojan design. Left: Trigger signal specification. Center: Logic-Rule-Driven (LRD) Trojans utilize discrete logic to validate signal features and duration (Td) before latching the activation signal. Right: Physical-Effect-Driven (PED) Trojans leverage component physics like rectification to extract baseband payloads from high-frequency inputs, effectively bypassing protection filters to manipulate sensor measurements.

1. Logic-Rule-Driven (LRD) Trojan

This design adds lightweight analog monitoring logic to the sensor.

  • Triggering: It uses high-impedance parallel monitoring to detect specific features like frequency (via high-pass filters and peak detectors) and amplitude (via window comparators).
  • Reliability: To prevent accidental activation from environmental noise, it includes a duration detector (using capacitor charge accumulation) that requires the trigger signal to persist for a specific time (e.g., 4 seconds) before the payload activates.
  • Payload: Once triggered, analog switches can force the output to a constant value (static spoofing) or induce a short circuit to ground (Permanent Denial-of-Service).

2. Physical-Effect-Driven (PED) Trojan

This design is even stealthier, avoiding explicit logic by tuning the sensor's native analog behaviors.

  • Mechanism: It exploits effects like amplifier saturation, rectification, and over-amplification. For example, by slightly modifying DC bias resistors, an attacker can cause an amplifier to clip asymmetrically when hit with a high-frequency signal, effectively "demodulating" a malicious low-frequency payload that bypasses the sensor's standard filters.
  • Stealth: Because it uses existing components or subtle parameter changes, these Trojans can often be mistaken for manufacturing defects or tolerances rather than malicious implants.

Datasets

While not a machine learning paper, the authors performed an extensive evaluation using 14 different commercial-off-the-shelf (COTS) sensors spanning six transduction principles:

  • Light: Adafruit GL5506, Vishay TEMT6000, Siemens QRB1.
  • Temperature: Amphenol NHQ103B, Heraeus PT100, People RETCB.
  • Microphone: CUI CMC-9745, DFRobot Sen0487.
  • Infrared: Murata IRA-S210, Adafruit SR501.
  • Pressure: Dayang DYTB-002.
  • Inclination: Analog ADXL335, ADXL337.

For stealth testing on microphones, they utilized the MUSAN audio corpus to simulate realistic household noise environments.

Key Insights and Ideas

  • Bypassing the "Digital" Mindset: Most hardware Trojan research focuses on digital logic (waiting for a specific bit pattern). SenTrojan proves that the analog front-end is equally vulnerable, using physical properties (frequency/amplitude) as the "key" to the backdoor.
  • Exploiting Safety Features: One of the most clever PDoS methods involves triggering a sensor's own thermal shutdown or overcurrent protection repeatedly, effectively locking the device in a permanent failure state.
  • Stealth through High Impedance: By designing the LRD monitoring circuit with extremely high input resistance, the authors ensured that the Trojan draws negligible current and does not load the sensing path, making it nearly invisible to standard performance benchmarks.
  • Frequency-Domain Backdoors: PED Trojans can use high-frequency carriers (e.g., 27 kHz ultrasound) to carry low-frequency commands. The sensor's own non-linearity acts as the demodulator, allowing the attack to bypass low-pass filters designed to protect the system.

Key Results

The evaluation across 14 sensors demonstrated that SenTrojan is both highly effective and remarkably stealthy:

  • High Success: Achieved an average Successful Trigger Rate (STR) above 95%.
  • Near-Perfect Stealth: Maintained over 97% similarity to benign sensor behavior during normal operation.
  • Zero False Alarms: The LRD designs achieved a 0% false-trigger rate even under continuous 72-hour monitoring in noisy environments.
  • Minimal Overhead: The Trojans introduced less than 5% average deviation in critical metrics like time drift, noise floor, and response time.
  • Real-World Impact: In a Hardware-In-the-Loop (HIL) simulation, a Trojan-compromised brake pressure sensor was used to cause a vehicle to collide with an obstacle by forcing the brake signal to zero upon receiving an ultrasonic trigger.

Limitations and Future Work

  • Defense via Shielding: The authors note that passive shielding (metal casings, sound-absorbing foam) can dampen the physical signals used to trigger the Trojans.
  • Detection Challenges: While LRD Trojans might be detectable via impedance-based monitoring (if a "golden model" is available), PED Trojans are much harder to catch because their component-level changes are so subtle.
  • Future Directions: The study suggests a need for new supply-chain verification techniques that look for "out-of-band" sensitivities and runtime anomaly detection that correlates multiple sensor inputs to spot inconsistencies caused by a triggered Trojan.

Summary updated

Abstract

Sensors bridge the physical and digital worlds, making their integrity critical to cyber-physical system security. Existing sensor attacks largely exploit unintended and sensor-specific physical vulnerabilities, limiting their generalizability and motivating a different threat model, i.e., whether an adversary can implant hidden attack surfaces into sensors before deployment. In this paper, we present SenTrojan, an analog hardware Trojan framework for sensors. SenTrojan embeds malicious functionality into analog signal circuitry commonly shared across diverse sensor architectures, and activates the hidden behavior through crafted external physical signals. Once activated, SenTrojan can induce sensor malfunctions ranging from permanent denial-of-service to static or dynamic spoofing. To achieve these effects under different architectural constraints, we introduce two complementary designs. The Logic-Rule-Driven (LRD) Trojan adds lightweight analog monitoring logic to detect trigger features such as frequency and amplitude, enabling reliable activation across various sensor types. The Physical-Effect-Driven (PED) Trojan targets sensors with exploitable native analog effects, e.g., nonlinearity and rectification, and realizes attacks with minimal component insertion. We evaluate SenTrojan on 14 sensors spanning 6 transduction principles and demonstrate its effectiveness across diverse real-world scenarios. Overall, SenTrojan achieves an average successful trigger rate above 95%, maintains over 97% similarity to benign behavior, incurs zero false-trigger rate for LRD designs, and introduces less than 5% average deviation in drift, noise floor, and response time. To our knowledge, this work is the first systematic study of analog sensor Trojans, broadening sensor security from unintended physical vulnerabilities to intentional supply-chain compromise.
BibTeX
@misc{wang2026sentrojan,
  author = {Kun Wang and Ruochen Zhou and Wenjun Zhu and Shilin Xiao and Kai Wang and Chen Yan and Xiaoyu Ji and Wenyuan Xu},
  title = {{SenTrojan: Revealing Analog Hardware Trojan Attacks on Sensors}},
  year = {2026},
  eprint = {2611.06103},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.06103},
}
Figure