PaperlayerView exact revision on arXiv

Source-linked AI summary

Analog Pin Directionality as an Exfiltration Attack Surface in Mixed-Signal ICs

Ramana Ranganatham, Chirag Adiga, Michael Zuzak, Tejasvi Das

arXiv:2609.19111v1cs.CRcs.AR

TL;DR

Specification-oriented testing may miss data-bearing perturbations on nominally input-only analog pins, leaving their directionality insufficiently verified as a security property. The paper models and demonstrates an offset-modulation attack in a PPG AFE; silicon measurements show recovery through the pin at rates up to 10 kbps. The results support treating analog pin directionality as a property for explicit verification and defense.

  • Problem

    Specification-oriented analog tests may not detect subtle, structured analog perturbations that remain within nominal operating bounds, leaving input-pin directionality insufficiently examined as a security property.

    Source: §I · p. 1

  • Method

    The paper models a Trojan payload that uses data-dependent circuit-offset modulation to exfiltrate through a nominally input-only pin, and identifies three enabling host conditions.

    Source: A. Goals and Contributions · p. 1

  • Results

    Silicon measurements demonstrate exfiltration through the input pin at bit rates up to 10 kbps and recover a 256-bit PRBS message at 6 bps with zero observed errors.

    Source: §V · p. 9

  • Takeaways & Limitations

    Analog pin directionality should be explicitly verified, covered in production testing, validated at system level, and defended at runtime rather than inferred from nominal signal flow.

    Source: §V · p. 9

Abstract

from arXiv · show

Mixed-signal SoCs rely on nominally input-only analog pins to acquire off-chip signals, but the directionality of these interfaces is generally treated as a functional property rather than explicitly verified as a security property. This work identifies and experimentally demonstrates a directionality-based class of analog and mixed-signal (AMS) exfiltration attacks in which data-dependent circuit-offset modulation converts a nominally input-only pin into an outbound information channel. We analytically model the attack mechanism and identify three enabling host conditions: a closed-loop amplifier, an exposed amplifier input, and sufficiently high impedance at that pin. This attack class is validated through a representative silicon case study using a photoplethysmography (PPG) analog front-end (AFE) fabricated in a commercial 55-nm CMOS process. The payload incurs $<$0.001\% area overhead relative to typical biosensing AFEs. Under the evaluated conditions, payload activation reduces the filtered PPG-output SNR by only 0.03~dB, while the maximum HT-induced perturbation of 5.9\% of the PPG amplitude remains within the 34.3\% benign variation at the exposed sensor-input pin across process and temperature. The raw exfiltration SINR remains below -20~dB, while targeted filtering increases it above 14~dB and enables signal recovery. Silicon measurements demonstrate data exfiltration through the input pin at bit rates up to 10~kbps and error-free recovery of a PRBS message. These results expose a conventional test-observability gap and establish analog pin directionality as an AMS security property requiring explicit verification, test coverage, and defense rather than being inferred from nominal signal flow.

I. INTRODUCTION

The paper identifies a security gap in specification-oriented analog testing and demonstrates an attack that turns a nominally input-only pin into an outbound channel. Its PPG AFE case study shows that the channel can carry data while host-function changes remain small under evaluated conditions.

A. Goals and Contributions

  • Data-dependent modulation of internal circuit offset turns a nominally input-only analog pin into an outbound channel when the host has three identified electrical conditions.The conditions are a closed-loop amplifier, an exposed amplifier input, and sufficiently high impedance at that pin.Source: A. Goals and Contributions · p. 1
  • 39.284 µm^2 payload area is <0.001% of typical biosensing AFEs, and the exposed sensor-input pin recovers a 256-bit pseudorandom test message.The payload is embedded in a PPG AFE fabricated in a commercial 55-nm CMOS process.Source: A. Goals and Contributions · p. 1
  • A 0.03 dB filtered PPG-output SNR change accompanies raw input-pin SINR below -20 dB, while targeted filtering raises exfiltration SINR above 14 dB.The informed receiver uses knowledge of the signaling band to recover transmitted data.Source: A. Goals and Contributions · pp. 1–2
  • Candidate host domains include radar, communication, wearable monitoring, biomedical sensing, and MEMS.The proposed attack class targets AMS interfaces rather than only the PPG case study.Source: A. Goals and Contributions · p. 1, Figure 1 · p. 2

I. INTRODUCTION

  • Specification-oriented tests do not expose the channel unless they monitor nominally input-only pins for data-bearing spectral activity.The introduction notes that conventional analog tests evaluate nominal directionality and specification compliance, while subtle structured perturbations can remain within operating bounds.Source: §I · p. 1, A. Goals and Contributions · p. 2

II. BACKGROUND, PROBLEM FORMULATION, AND THREAT MODEL

This section situates analog hardware Trojans within broader hardware-security concerns and formulates a question about covert exfiltration through analog inputs. It highlights the limits of specification-driven testing for continuous analog behavior.

A. Background

  • Hardware Trojans can infiltrate functionality or exfiltrate confidential data, and distributed design and manufacturing expose multiple lifecycle stages to Trojan insertion.Small malicious blocks may be difficult to identify during routine verification or visual inspection.Source: A. Background · p. 2
  • Analog and mixed-signal security is comparatively underexplored, while prior analog Trojan work uses coupling, emissions, or parameter shifts to subvert systems.The growing use of analog functionality motivates study of analog-domain payloads that exploit existing infrastructure for covert exfiltration.Source: A. Background · p. 2

B. Problem Formulation

  • Analog circuits operate through continuous-time behavior governed by bias, impedance, and waveforms, allowing subtle parameter perturbations without explicit logic events.Such effects can remain within specifications and be difficult to distinguish from process, voltage, and temperature variation.Source: B. Problem Formulation · p. 2
  • Specification-driven analog tests target violations and selected faults rather than exhaustive behavior coverage, with production and characterization constrained by practical costs and observability.The paper asks whether attacker-controlled data can be encoded as offset perturbations at a nominally input-only pin while preserving host function and evading tests that do not monitor that pin.Source: B. Problem Formulation · p. 2
  • Table I is identified as a comparison of representative covert and exfiltration mechanisms in prior work.The supplied table caption establishes its scope but not its entries or comparisons.Source: Table I · p. 3

C. Related Work

Prior covert-channel and detection work does not establish coverage of low-amplitude analog-state modulation on nominally input-only pins. The threat model treats secret extraction and payload activation as existing capabilities and focuses on transport through an exposed analog input.

C. Related Work

  • Prior channels use RF, electromagnetic, thermal, power, acoustic, optical, peripheral, and software pathways, but coverage of input-pin analog-state modulation is not established.Existing detection approaches include statistical analysis and machine learning; the passage says they do not inherently observe this channel.Source: C. Related Work · p. 3

D. Threat Model

  • The threat model assumes a supplied secret-bearing bitstream and a payload inserted by an untrusted foundry or malicious design-stage actor; extraction and activation are modular capabilities outside this work’s transport focus.The attacker is assumed to know the signaling band and receiver processing and to have physical access to the exposed sensor-input net for high-impedance probing and targeted recovery.Source: D. Threat Model · p. 3
  • The evaluated attack does not require changes to normal digital I/O, packaging, test infrastructure, or on-chip cryptographic operations, but assumes no targeted monitoring of input pins for outbound modulation.Success means recovering the bitstream while the host AFE satisfies the evaluated functional specifications.Source: D. Threat Model · p. 3

III. ANALOG INPUT-PIN EXFILTRATION METHODOLOGY

  • Candidate hosts combine a closed-loop amplifier, an exposed amplifier input, and sufficiently high pin impedance; the methodology analyzes insertion targets, topology, and operating conditions.The passage names ECG, PPG, EEG, off-chip temperature sensing, and optical communication interfaces as possible examples.Source: §III · p. 3

A. Host-Interface and Pin-Selection Analysis

The analysis favors an exposed, high-impedance sensor input as an exfiltration node in a typical current-mode AFE. The Trojan reuses the amplifier’s differential-pair devices to encode leaked bits as offset modulation at that pin.

A. Host-Interface and Pin-Selection Analysis

  • Under the stated impedance assumptions, the sensor pin has greater effective impedance than the reference pin, so Vsns is favored for payload insertion and data transport.The analysis assumes a low-impedance reference source and a high-impedance sensor node; zsns≈zsensor∥zfb and zref≈rref.Source: A. Host-Interface and Pin-Selection… · p. 4

B. Trojan Insertion

  • The payload adds switched MP and MN transistor fingers to amplifier U1, with each switch implemented as a single NMOS transistor to reduce area.The topology can use analog dummies as Trojan transistors to further reduce area overhead, and assumes leaked data reaches the payload through a subtle modification of existing interconnect.Source: B. Trojan Insertion · p. 4, Figure 3 · p. 4
  • Leaked bits control the switches, adding differential-pair fingers that induce amplifier offset and create an analog-domain DAC and ASK modulator without routing data through the host ADC or a dedicated DAC.The output-offset mapping depends on added transistor fingers and the Trojan-transistor W/L ratios.Source: B. Trojan Insertion · p. 4
  • The HT-induced offset must exceed input-referred circuit noise for recovery while remaining well below the sensor signal to limit host-AFE impact.Payload transistor sizing and exfiltration frequency can prioritize stealth or speed; reliable operation requires a closed-loop amplifier, an exposed input, and sufficiently high pin impedance.Source: B. Trojan Insertion · p. 4

C. Channel Characteristics

The section models recoverability and capacity for the covert channel using receiver SINR, bit-error rate, and Shannon’s theorem. Usable transmission bands depend on the host and operating conditions.

C. Channel Characteristics

  • Targeted filtering in the known signaling band can increase recovered SINR, while payload-frequency selection trades recoverability against detectability.The attacker may perform filtering in software after measurement, reducing required receiver hardware.Source: C. Channel Characteristics · p. 5
  • For ASK modulation with Gaussian-distributed noise, the HT’s SINR determines the corresponding bit error rate.The recoverability expression uses the RMS sensor signal over the same receiver bandwidth as the input-referred offset and noise.Source: C. Channel Characteristics · p. 5
  • Channel capacity is estimated with Shannon’s theorem across frequency, with amplifier bandwidth bounding covert-channel bandwidth.The input-pin transfer response, host-signal spectrum, and recoverable SINR determine usable transmission bands, so capacity depends on the host and operating conditions.Source: C. Channel Characteristics · p. 5

1) Gaps in Current AMS Test Methods:

Conventional AMS testing targets specification compliance and may not observe low-SINR, data-dependent modulation on nominally input-only pins. The section defines a relative detectability index and motivates security-oriented spectral monitoring.

1) Gaps in Current AMS Test Methods:

  • Conventional AMS tests do not inherently provide continuous, high-resolution observation of every input pin in mission mode, leaving low-SINR modulation outside routine specification screening.Probe loading, measurement limits, and the cost of resolving perturbations against process and temperature variation further constrain validation; the channel is observable when tests monitor data-dependent spectral activity with sufficient bandwidth and resolution.Source: 1) Gaps in Current AMS Test Methods · p. 5

2) Relative Detectability:

  • The relative detectability index Mdet compares payload operating points using host SNR variation and signal frequencies.Its terms account for HT-induced and total SNR variation, measurement uncertainty, and frequency separation from nominal host activity; it is not a detection probability or specific detector’s performance.Source: 2) Relative Detectability · pp. 5–6
  • Nominally input-only pins must be treated as potential outbound interfaces and monitored during normal operation for data-dependent modulation.The proposed defensive capability requires sufficient bandwidth and resolution, motivating security-oriented spectral testing under representative operating conditions.Source: 2) Relative Detectability · p. 6

IV. SILICON CASE STUDY: PPG FRONT-END

The silicon case study uses a fabricated PPG front end to examine the attack’s impact and observability under conventional testing. The measurement setup characterizes a 500 Hz test tone and validates transmission with a PRBS.

IV. SILICON CASE STUDY: PPG FRONT-END

  • The PPG AFE architecture comprises a photodetector-bias current source, TIA, anti-aliasing LPF, ADC, and digital processing block.The case study maps the sensor current to Isensor=IPD1, the amplifier output to Vamp=VPPG, and the exposed pin to Vpd=Vsns.Source: §IV · p. 6
  • The embedded payload occupies 39.284 µm2, or 0.8% of the front-end amplifier area and less than 0.001% of typical biosensing AFE area.The chip was fabricated in a commercial 55-nm CMOS process and includes a TIA and adjustable photodiode-bias current source.Source: §IV · p. 6

A. Measurement Setup

  • The fabricated AFE and embedded payload were characterized on a mixed-signal measurement bench, shown alongside the chip micrograph in Fig. 5.The bench includes a photodiode, red LED, microcontroller, and test PC.Source: Figure 5 · p. 6, A. Measurement Setup · p. 6
  • Measurements used 140 mV/µA transimpedance gain, a 1 Hz PPG frequency, and a 100 Hz external output LPF.The channel was characterized with a continuous 500 Hz tone extracted from Vpd using a 150 Hz high-pass filter, then validated with PRBS transmission; measurements were at room temperature.Source: A. Measurement Setup · p. 6

B. Attack Validation

Silicon measurements show that the payload preserves the PPG output while its input-pin signal is recoverable with targeted filtering. Simulations and measurements characterize its robustness, offset control, and PRBS transmission.

B. Attack Validation

  • Across evaluated process corners and temperatures, the maximum HT perturbation of 5.9% of PPG amplitude remains within the 34.3% benign input-voltage variation.Conventional specification-based screening would not flag the payload through an out-of-spec result under these conditions.Source: B. Attack Validation · pp. 6–7, Figure 6 · p. 7
  • Across 1000 mismatch simulations, most samples had BER below 30×10^-6, although the upper tail showed substantial degradation for some samples.The reported BER distribution is estimated across Monte Carlo mismatch simulations.Source: Figure 7 · p. 7, B. Attack Validation · p. 7
  • One and two additional HT fingers produce measured input-offset shifts of 3.7 mV and 6.8 mV; at 500 Hz without PPG, one finger yields a modeled minimum BER of 6.7 × 10^-6.Further measurements use one finger to limit host perturbation and detectability; additional fingers reduce BER but increase perturbation.Source: B. Attack Validation · p. 7, Figure 8 · p. 7
  • A 256-bit PRBS message was recovered with zero errors at 6 bps using a 500 Hz carrier and digital high-pass filtering.The receiver used a 250 Hz-corner digital high-pass filter to extract and demodulate the message.Source: B. Attack Validation · p. 7, Figure 10 · p. 8
  • Payload activation changes filtered PPG-output SNR by only 0.03 dB.The result is reported under the evaluated operating conditions.Source: Figure 11 · p. 8, B. Attack Validation · p. 8
  • Raw input-pin exfiltration SINR remains below -20 dB, while targeted filtering raises it above 14 dB and enables recovery.The raw signal is beneath the interference-plus-noise floor, preventing reliable recovery without filtering.Source: Figure 12 · p. 8, B. Attack Validation · p. 8

C. Tradeoffs and Comparisons

The measured operating points expose a tradeoff between signal recovery and detectability as payload frequency changes. The work also reports channel capacity and compares demonstrated transmission rates with representative prior covert channels.

C. Tradeoffs and Comparisons

  • Greater spectral separation improves SINR and BER but increases detectability; in the 100 Hz-to-1 kHz band, estimated BER ranges from 1.66 × 10^-2 to 2.51 × 10^-2.The measured relationship was evaluated with the PPG signal present.Source: C. Tradeoffs and Comparisons · p. 8
  • The relative-detectability index ranges from 1.40 × 10^-3 at 500 mHz to 27.81 at 10 kHz; computed channel capacity up to 10 kHz is 30.19 kbps.Capacity was computed from measured values using the trapezoidal rule.Source: C. Tradeoffs and Comparisons · p. 8
  • Alternating-pattern measurements support 0.1 bps to 10 kbps, while a 256-bit PRBS message was recovered with zero observed errors at 6 bps; cited prior channels support 1 bps to 10.21 kbps with BERs ≤23.4%.The rate range is based on measured SINR with BER estimated; the PRBS transmission is a separate validation.Source: C. Tradeoffs and Comparisons · p. 8

V. CONCLUSIONS

The conclusions identify a directionality-based AMS exfiltration attack in which data-dependent offset modulation turns a nominally input-only pin into an outbound channel. The silicon case study demonstrates low host impact and recoverability, motivating explicit security verification of analog pin directionality.

V. CONCLUSIONS

  • The attack uses data-dependent circuit-offset modulation to turn a nominally input-only analog pin into an outbound channel, with candidate hosts requiring a closed-loop amplifier, exposed input, and sufficiently high pin impedance.The case study embeds an analog-domain HT payload in a PPG AFE fabricated in commercial 55-nm CMOS.Source: §V · p. 9
  • The payload adds <0.001% area relative to typical biosensing AFEs; measured filtered PPG-output SNR changes by 0.03 dB, while targeted filtering raises input-pin SINR from below -20 dB to above 14 dB.A 256-bit PRBS message was recovered at 6 bps with zero observed errors, and alternating-pattern measurements support rates up to 10 kbps.Source: §V · p. 9
  • Analog pin directionality should be treated as a security property requiring explicit verification, production-test coverage, system-level validation, and runtime defense.The conclusion cautions against inferring security from nominal signal flow.Source: §V · p. 9
Loading 2609.19111v1…