Terminal Brain Damage: Exposing the Graceless Degradation in Deep Neural Networks Under Hardware Fault Attacks

@article{Hong2019TerminalBD,
  title={Terminal Brain Damage: Exposing the Graceless Degradation in Deep Neural Networks Under Hardware Fault Attacks},
  author={Sanghyun Hong and Pietro Frigo and Yigitcan Kaya and Cristiano Giuffrida and Tudor Dumitras},
  journal={ArXiv},
  year={2019},
  volume={abs/1906.01017}
}
  • Sanghyun Hong, Pietro Frigo, +2 authors Tudor Dumitras
  • Published 2019
  • Computer Science
  • ArXiv
  • Deep neural networks (DNNs) have been shown to tolerate "brain damage": cumulative changes to the network's parameters (e.g., pruning, numerical perturbations) typically result in a graceful degradation of classification accuracy. However, the limits of this natural resilience are not well understood in the presence of small adversarial changes to the DNN parameters' underlying memory representation, such as bit-flips that may be induced by hardware fault attacks. We study the effects of… CONTINUE READING

    Citations

    Publications citing this paper.
    SHOWING 1-10 OF 12 CITATIONS

    DeepHammer: Depleting the Intelligence of Deep Neural Networks through Targeted Chain of Bit Flips

    VIEW 5 EXCERPTS
    CITES METHODS, RESULTS & BACKGROUND
    HIGHLY INFLUENCED

    TBT: Targeted Neural Network Attack with Bit Trojan

    VIEW 6 EXCERPTS
    CITES BACKGROUND
    HIGHLY INFLUENCED

    TRRespass: Exploiting the Many Sides of Target Row Refresh

    VIEW 3 EXCERPTS
    CITES BACKGROUND

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 77 REFERENCES

    Fault injection attack on deep neural network

    VIEW 4 EXCERPTS
    HIGHLY INFLUENTIAL

    Practical Fault Attack on Deep Neural Networks

    VIEW 4 EXCERPTS
    HIGHLY INFLUENTIAL

    Ares: A framework for quantifying the resilience of deep neural networks

    VIEW 3 EXCERPTS

    Hardware Trojan Attacks on Neural Networks

    VIEW 2 EXCERPTS