Kahraman D. Akdemir

Design of Cryptographic Devices Resilient to Fault Injection Attacks Using Nonlinear Robust Codes

This chapter mainly discusses robust and partially robust codes and their application to various cryptographic primitives. Initially, robust nonlinear codes are described in detail and their error detection capabilities are measured theoretically. Next, various nonlinear constructions are provided and their potential applications are described. More specifically, we discuss the protection of the AES data path, finite state machines (FSMs), and elliptic curve cryptosystems (ECCs). The main advantage of robust codes is that they are nonlinear and hence the success of an injected fault is data-dependent. As a result, error detection using nonlinear robust codes is one of the most effective solutions to active fault injection attacks.

Novel PUF-Based Error Detection Methods in Finite State Machines

We propose a number of techniques for securing finite state machines (FSMs) against fault injection attacks. The proposed security mechanisms are based on physically unclonable functions (PUFs), and they address different fault injection threats on various parts of the FSM. The first mechanism targets the protection of state-transitions in a specific class of FSMs. The second mechanism addresses the integrity of secret information. This is of particular interest in cryptographic FSMs which require a secret key. Finally, the last mechanism we propose introduces a new fault-resilient error detection network (EDN). Previous designs for EDNs always assume resilience to fault injection attacks without providing a particular construction. The PUF-based EDN design is suitable for a variety of applications, and is essential for most fault resilient state machines. Due to the usage of PUFs in the proposed architectures, the state machine will enjoy security at the logical level as well as the physical level.

An emerging threat: eve meets a robot

In this work, we study the emerging security threats in a quickly proliferating field: robotics. The next generation robots embody most of the networking and computing components we normally use for everyday computing. Thus, the next generation robots virtually inherit all of the security weaknesses we are struggling with today. To make things worse, vulnerabilities in robots are much more significant, as they physically interact with their surroundings which include human beings. In this paper, we first provide a classification of potential physical attacks on robots. In addition, we outline a concrete active attack and propose a countermeasure.

Non-linear Error Detection for Finite State Machines

We propose the use of systematic nonlinear error detection codes to secure the next-state logic of finite state machines (FSMs). We consider attacks under an adversarial model which assumes an advanced attacker with high temporal and spatial fault injection capability. Due to the non-uniform characteristics of FSMs, simple application of the systematic non-linear codes will not provide sufficient protection. As a solution to this problem, we use randomized masking. Furthermore, we show that our proposal detects injected faults with probability exponentially close to 1.

Non-linear error detection for elliptic curve cryptosystems

The authors propose applying systematic non-linear error-detection codes to protect elliptic curve point addition and doubling operations against active fault attacks. These codes provide nearly perfect error-detection capability (except with exponentially small probability) at reasonable overhead. The proposed technique is applied to secure point addition and doubling operations for both Weierstrass and Edwards curves using different coordinate systems (i.e. affine and projective). The authors observe that the Weierstrass-based elliptic curve systems can be protected with reasonable area overhead. However, due to its balanced normal form, Edwards formulation is more appropriate for the non-linear error-detection technique proposed here. In addition, the proposed technique is compared with the method discussed by Gaubatz et al. (2006), where an error-detection technique is proposed for robust public key arithmetic. When compared with their method, the proposed technique provides approximately the same level of security with much less overhead. For Edwards curves, the overhead of the proposed scheme is less than half (42–46%) of the overhead of scheme proposed by Gaubatz et al. (2006). In addition, the overhead of the proposed scheme is 52–81% of the overhead of scheme proposed by Gaubatz et al. (2006) for different versions of the Weierstrass curves.