Fault Analysis in Cryptography

Stream ciphers are highly frequent in the realm of secret key cryptography because of their low-computational complexity and latency. Though variety of attack models for testing the robustness of various stream ciphers are available, the differential fault analysis (DFA) model is the most popular. DFA is used not only to disclose the identity of any unknown cipher but also to recover the secret KEY and the cipher’s internal state. Deep learning approaches extract patterns from large amounts of cryptographic data, which aids in crypt-analysis. Here, single bit and multi-bit faults are injected into the internal register of trivium cipher, and artificial neural network (ANN) is used to identify the injected fault location without any handcrafted features of keystream unlike other existing state-of-the-art fault analysis techniques. In order to enhance the speed of fault analysis, we have used system on chip (SoC) platform where ANN-based classifier has been placed in the programmable logic area as an accelerator. This enhances the speed of fault analysis by 9% for single bit fault, 12% for two adjacent bit faults, and 13% for three adjacent bit faults compared to their software models. To the best of the author’s knowledge, this is the first of kind research where ANN-based hardware accelerator has been used for fault location identification on SoC platform.

Differential fault analysis (DFA) has already been applied to attack many block ciphers with the help of inducing some faults at the last few rounds of block ciphers. Currently, a general countermeasure against DFA is to protect the last few rounds of block ciphers by means of redundancy. In this paper, we present a new fault attack on block ciphers called linear fault analysis (LFA), in which linear characteristics for some consecutive rounds of a block cipher will be utilized instead of exploiting differential distributions of S-Boxes within the block cipher in DFA. Basically, the new approach can handle the case that faults are induced several rounds earlier compared to DFA, thus leading to a threat to the protected implementations (against DFA) of block ciphers. For the purpose of illustration, we mount an effective attack on SERPENT by adopting LFA and achieve a good cryptanalytic result on SERPENT. We hope that our work enriches the picture on the applicability of fault attacks to block ciphers and could be beneficial to the security evaluation of block ciphers.

This paper proposes a fault analysis technique on LED by combining algebraic cryptanalysis and differential fault analysis (DFA). The technique is called algebraic differential fault analysis (ADFA). In ADFA on LED, we use DFA to deduce the possible fault differences of the correct and faulty S-Box input in the last round, and convert them into algebraic equations. We then combine the equation set of LED with the injected fault and use the CryptoMiniSat solver to recover the secret key. Our experiments show that, on a common PC, ADFA can succeed on LED under the nibble-based fault model within three minutes and with only one fault injection, which is more efficient than previous DFA work. To evaluate DFA on LED, we first propose an improved evaluation algorithm of DFA, then provide a modified ADFA approach to compute the solutions for the secret key. The results are more accurate than previous work. We also successfully extend ADFA on LED to other fault models using a single fault injection, where traditional DFAs are difficult to launch.

Side channel analysis(SCA) focuses on attacking cryptographic algorithm implementations by collecting leaked information and analyzing the correlation between the leaked and key information in the cryptographic system. Currently, SCA is the main threat to cryptographic implementations, with DPA (differential power analysis) and DFA (differential fault analysis) being two of the most threatening types of SCA. However, the existing countermeasures against DPA and DFA have shortcomings and can hardly protect cipher designs perfectly. Based on a comprehensive analysis of DPA and DFA, this paper proposes a new idea of using an evolutionary cipher (EVOC) against DPAs based on the model proposed by Kocher, and several different new types of DFAs. In fact, an EVOC is a kind of dynamic cipher designing method. Moreover, combined with intelligent searching algorithms and cryptography design policies, EVOCs utilize the dynamic and unpredictable properties of TRNG (a truly random number generator) to ensure cipher design at a high level of security. Therefore, with an EVOC we could design cryptography algorithms with high security to resist mathematical analysis as well as DPAs and DFAs. This paper analyzes and proves the security, efficiency, cost, and original features of EVOCs against DPA and DFA both theoretically and experimentally. Compared with existing countermeasures against SCAs, EVOCs use dynamic non-linear operations to destroy the bases of DPA and DFA. At the algorithm level, EVOCs are more efficient than other resistance methods. Based on the theory and experiments, this paper proposes certain modifications to the EVOC in the AES (advanced encryption standard) algorithm, which can be referenced by other dynamic designs of cryptographic algorithms.

Block ciphers are typically resistant to direct attacks, such as an exhaustive key search or cryptanalysis, all of which require too many resources to achieve an efficient attack. Many block ciphers are examined for their resistance to less direct attacks that target a given implementation. Of these attacks, fault attacks are amongst the most effective at retrieving information on secret key, and require specific countermeasures to be included in an implementation. In this paper we describe a simple platform for the study of fault injection and analysis in the context of fault attacks block ciphers based on a Feistel structure (e.g. DES). We show that an attacker who can successfully inject faults into a block cipher can reduce the complexity of an attack to derive the secret key. We also present a novel version of the attack than can be applied to Triple-DES by independently injecting faults in the second and third instantiations of DES involved in a Triple- DES computation.

Differential fault analysis on the contracting UFN structure, with application to SMS4 and MacGuffin

The development of differential fault analysis (DFA) techniques and mechanisms to inject faults into cryptographic circuits brings with it the need to use protection mechanisms that guarantee the expected level of security. The AES cipher, as a standard, has been the target of numerous DFA techniques, where its security has been compromised through different formulations and types of fault injections. These attacks have shown vulnerabilities of different AES implementations and building blocks. Consequently, several solutions have been proposed that provide additional protection to cover the identified vulnerabilities. In this paper, an extensive analysis has been carried out covering the existing fault injection techniques, the types of faults, and the requirements needed to apply DFA. Additionally, an analysis of the countermeasures reported in the literature is also presented, considering the protection provided, the type of faults considered, and the coverage against fault attacks. The eight different types of fault that allow us to perform DFAs on the AES cipher have been differentiated, as well as the vulnerabilities of the cipher. On the other hand, two comparisons have been made considering frequency penalty vs. area and fault coverage vs. area and frequency overhead. A metric has been proposed to compare the fault coverage of all the proposed solutions. To conclude, a final analysis is presented discussing the key aspects when choosing a particular solution and the possible development of new countermeasures to provide further protection against DFA.

While modern masking schemes provide provable security against passive side-channel analysis (SCA), such as power analysis, single faults can be employed to recover the secret key of ciphers even in masked implementations. In this paper, we propose random space masking (RS-Mask) as a countermeasure against both power analysis and statistical fault analysis (SFA) techniques. In the RS-Mask scheme, the distribution of all sensitive variables, faulty and/or correct values is uniform, and it therefore protects the implementations against any SFA technique that exploits the distribution of intermediate variables, including fault sensitivity analysis (FSA), statistical ineffective fault analysis (SIFA) and fault intensity map analysis (FIMA). We implement RS-Mask on AES, and show that a SIFA attack is not able to identify the correct key. We additionally show that an FPGA implementation of AES, protected with RS-Mask, is resistant to power analysis SCA using Welch's t-test. The area of the RSMasked AES is about 3.5 times that of an unprotected AES implementation of similar architecture, and about 2 times that of a known FPGA SCA-resistant AES implementation. Finally, we introduce infective RS-Mask that provides security against differential techniques, such as differential fault analysis (DFA) and differential fault intensity analysis (DFIA), with a slight increase in overhead.

Basically, Differential Fault Analysis (DFA) against ciphers consists of two stages: fault induction and fault exploitation. Success rate of the latter strongly depends upon the availability of the required type of some faulty ciphertexts, which typically assumes that adversaries have precise control over the location, timing and/or even the type of the fault induced. In view of this, there seems to be a technical gap between these two stages. In this paper, by amplifying the fault propagation pattern inherent in block ciphers, we propose an algorithmic method to narrow this gap, or to relax the underlying assumption. Our method provides a faulty ciphertext filtering process between two basic stages in DFA. This additional process equips adversaries with the capabilities to deterministically decide whether or not one faulty ciphertext is relevant to a specific DFA, before or without performing DFA itself. We take AES as a concrete case of study and conduct theoretical analysis and simulation experiments as well, the results of which strongly demonstrate the validity and power of our proposed filtering method.

Differential Fault Attacks (DFA) on stream ciphers have been an active field of research. However, their practical realizations have not been reported in the public literature. Hence, the assumptions on the fault models made in the context of DFA for stream ciphers have not been studied. Furthermore, there have been few efforts reported on the popular stream cipher candidate, Grain-128. We consider a simple low-cost fault injection set-up, using clock glitches and show that in stream ciphers the critical path of the circuit affects few bit positions (the feedback bit for the Shift Registers in the stream ciphers). Thus the fault is often localized to single bit position, and because of the absence of required faulty ciphers makes existing theoretical DFAs invalid. In order to create multiple instance of faults, we use clock glitches to induce the fault, and then use the shifting property of the internal registers of Grain to create multiple instances of contiguously located faults. In parallel, we also develop a more relaxed DFA for Grain-128, to show that when the fault is k neighbourhood bits, k e {1,..,5}, the attack is successful to retrieve the key without knowing the locations or exact number of bits flipped by the internal fault. We also devise a technique for rejecting the bad faults with high probabilities, i.e., when the faults are not in the contiguous location as required in the attack. Combining the above attacks we demonstrate using a simple set-up via clock glitches that such faults can be practically obtained and analysed using the proposed attack algorithm to retrieve the key.

Algebraic fault analysis (AFA), which combines algebraic cryptanalysis with fault attacks, has represented serious threats to the security of lightweight block ciphers. Inspired by an earlier framework for the analysis of side-channel attacks presented at EUROCRYPT 2009, a new generic framework is proposed to analyze and evaluate algebraic fault attacks on lightweight block ciphers. We interpret AFA at three levels: the target, the adversary, and the evaluator. We describe the capability of an adversary in four parts: the fault injector, the fault model describer, the cipher describer, and the machine solver. A formal fault model is provided to cover most of the current fault attacks. Different strategies of building optimal equation set are also provided to accelerate the solving process. At the evaluator level, we consider the approximate information metric and the actual security metric. These metrics can be used to guide adversaries, cipher designers, and industrial engineers. To verify the feasibility of the proposed framework, we make a comprehensive study of AFA on an ultra-lightweight block cipher called LBlock. Three scenarios are exploited which include injecting a fault to encryption, to key scheduling, or modifying the round number or counter. Our best results show that a single fault injection is enough to recover the master key of LBlock within the affordable complexity in each scenario.

Literature on Differential Fault Analysis (DFA) on AES-128 shows that it is more difficult to attack AES when the fault is induced in the key schedule, than when it is injected in the intermediate states. Recent research shows that DFA on AES key schedule still requires two faulty cipher texts, while it requires only one faulty cipher text and a brute-force search of $2^8$ AES-128 keys when the fault is injected inside the round of AES. The present paper proposes a DFA on AES-128 key schedule which requires only one single byte fault and a brute-force search of $2^8$ keys, showing that a DFA on AES key schedule is equally dangerous as a fault analysis when the fault is injected in the intermediate state of AES. Further, the fault model of the present attack is a single byte fault. This is more realistic than the existing fault model of injecting three byte faults in a column of the AES key which has a less chance of success. To the best of our knowledge the proposed attack is the best known DFA on AES key schedule and requires minimum number of faulty cipher text. The simulated attack, running on 3GHz Intel Core 2 Duo desktop machine with 2GB RAM, takes around 35 minutes to reveal the secret key.

Physical attacks, such as fault attacks, pose a decisive threat for the security of devices in the Internet of Things. An important class of countermeasures for fault attacks is fault tolerant software that is applicable for systems based on COTS hardware. In order to evaluate software countermeasures against fault attacks, fault injection is needed. However, established fault injection approaches require manufactured products or hardware details (e.g. netlists, RTL models), which are not available when using COTS hardware. In this paper, we present a QEMU-based fault injection platform that supports commercial COTS processors that are widely-used in the embedded domain. This framework allows a system-level analysis of software countermeasures by featuring the simulation of high-level hardware faults targeting, for example, memory cells, register cells, or the correct execution of instructions. The framework supports the generation of realistic fault attack scenarios. We illustrate the practicability of the approach by presenting two exemplary use cases.

We present a fault analysis study of the ChaCha and Salsa families of stream ciphers. We first show that attacks like differential fault analysis that are common in the block cipher setting are not applicable against these families of stream ciphers. Then we propose two novel fault attacks that can be used against any variant of the ciphers. We base our attacks on two different fault models: the stuck-at fault model and the biased fault model. Each of them is exploited differently by the attacker. If the attacker knows the plaintexts and the ciphertexts both fault models can be successfully exploited. If the ciphers operate on fixed yet unknown plaintexts only the biased fault model can be successfully exploited. We evaluate exemplary attacks using both models in simulation. Their low complexity confirms that they are practical. To the best of our knowledge these are the first fault attacks against ChaCha and Salsa that do not require faults in the control flow (e.g. instruction skip).

Fault injection attacks (FIAs) are becoming a serious threat to the security of integrated circuits (ICs). Circuit designers need a simple and effective method to evaluate the ability of their IC designs against the FIAs at design stage. To address the need, this paper based on FPGA emulation proposes a method, which mimics FIAs on a circuit design implemented on a FPGA by instantiating a uniform architecture which can generate different fault models and span the space of faults and inserting fault injection logic into the design. After the circuit design is submitted to the emulation platform, the fault generation and injection are executed automatically, and a report about the ability of the IC design against the FIAs is generated. The method works at the circuit netlist level, leaves the source codes of the circuit design unmodified, and makes the evaluation a transparent process to the designers. The experiment with various fault models and differential fault analysis on AES-128 encryption circuit shows the effectiveness of the method.