Energy-Latency Attacks: A New Adversarial Threat to Deep Learning

Hanene F. Z. Brachemi Meftah, Wassim Hamidouche, Sid Ahmed Fezza, Olivier Déforges · 2025

Read Energy-Latency Attacks: A New Adversarial Threat to Deep Learning by Hanene F. Z. Brachemi Meftah, Wassim Hamidouche, Sid Ahmed Fezza, Olivier Déforges (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-04963.txt.

Description

Meftah and co-authors survey in 2025 a new kind of attack on deep learning systems: tricks that waste energy and slow replies. Big neural networks already draw much power, so makers build lean hardware and models that adapt to each task. Those lean designs open a fresh weak point.

An attacker can craft inputs that force the worst case path, draining power and time until the service stalls. The survey sorts known attacks with a classic scheme for hostile inputs. It also reviews how success is measured and what defenses exist. The authors aim to guide safer design of lean AI systems.

Catalog record

CREATORHanene F. Z. Brachemi Meftah, Wassim Hamidouche, Sid Ahmed Fezza, Olivier Déforges
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE105 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-04963

Browse the full library · About the Tower of Babel