Description
With the widespread development of artifical intelligence, Deep Neural Networks (DNN) have become valuable intellectual property (IP). In the past few years, software and hardware-based attacks targetting at the weights of the DNN have been introduced allowing potential attacker to gain access to a near-perfect copy of the victim's model. However, these attacks either fail against more complex architecture or loose in fidelity regarding the targeted model. In this talk, I will present a new attack methodology combining side-channels attacks with software attacks in order to extract more complex DNN. First, I will present the main methodologies from the state of the art, and the comparison that can be made with traditional cryptanalytic attacks. I will then describe how side-channel attacks can help with the current limitations and introduce our framework as well as our experimental results. This talk will be concluded by the presentation of some potentials future work.
Practical infos
Next sessions
-
Advanced techniques for fault injection attacks on integrated circuits
Speaker : Paul Grandamme - Laboratoire Hubert Curien, Université Jean Monnet
The security of integrated circuits is evaluated through the implementation of attacks that exploit their inherent hardware vulnerabilities. Fault injection attacks represent a technique that is commonly employed for this purpose. These techniques permit an attacker to alter the nominal operation of the component in order to obtain confidential information. Firstly, we propose the utilisation of[…]-
SemSecuElec
-
Fault injection
-
-
Anomalies Mitigation for Horizontal Side Channel Attacks with Unsupervised Neural Networks
Speaker : Gauthier Cler - SERMA Safety & Security
The success of horizontal side-channel attacks heavily depends on the quality of the traces as well as the correct extraction of interest areas, which are expected to contain relevant leakages. If former is insufficient, this will consequently degrade the identification capability of potential leakage candidates and often render attacks inapplicable. This work assess the relevance of neural[…]-
SemSecuElec
-
Side-channel
-
Machine learning
-
-
Side-Channel Based Disassembly on Complex Processors: From Microachitectural Characterization to Probabilistic Models
Speaker : Julien Maillard - CEA
Side-Channel Based Disassembly (SCBD) is a category of Side-Channel Analysis (SCA) that aims at recovering information on the code executed by a processor through the observation of physical side-channels such as power consumption or electromagnetic radiations. While traditional SCA often targets cryptographic keys, SCBD focuses on retrieving assembly code that can hardly be extracted via other[…]-
SemSecuElec
-
Side-channel
-
Hardware reverse
-