Conception et test des circuits et systèmes numériques à haute fiabilité et sécurité
Abstract
Research activities I carried on after my nomination as Chargé de Recherche deal with the definition of methodologies and tools for the design, the test and the reliability of secure digital circuits and trustworthy manufacturing. More recently, we have started a new research activity on the test of 3D stacked Integrated CIrcuits, based on the use of Through Silicon Vias. Moreover, thanks to the relationships I have maintained after my post-doc in Italy, I have kept on cooperating with Politecnico di Torino on the topics related to test and reliability of memories and microprocessors.
Secure and Trusted Devices
Security is a critical part of information and communication technologies and it is the necessary basis for obtaining confidentiality, authentication, and integrity of data. The importance of security is confirmed by the extremely high growth of the smart-card market in the last 20 years. It is reported in "Le monde Informatique" in the article "Computer Crime and Security Survey" in 2007 that financial losses due to attacks on "secure objects" in the digital world are greater than $11 Billions. Since the race among developers of these secure devices and attackers accelerates, also due to the heterogeneity of new systems and their number, the improvement of the resistance of such components becomes today’s major challenge.
Concerning all the possible security threats, the vulnerability of electronic devices that implement cryptography functions (including smart cards, electronic passports) has become the Achille’s heel in the last decade. Indeed, even though recent crypto-algorithms have been proven resistant to cryptanalysis, certain fraudulent manipulations on the hardware implementing such algorithms can allow extracting confidential information. So-called Side-Channel Attacks have been the first type of attacks that target the physical device. They are based on information gathered from the physical implementation of a cryptosystem. For instance, by correlating the power consumed and the data manipulated by the device, it is possible to discover the secret encryption key. Nevertheless, this point is widely addressed and integrated circuit (IC) manufacturers have already developed different kinds of countermeasures.
More recently, new threats have menaced secure devices and the security of the manufacturing process. A first issue is the trustworthiness of the manufacturing process. From one side, secure devices must assure a very high production quality in order not to leak confidential information due to a malfunctioning of the device. Therefore, possible defects due to manufacturing imperfections must be detected. This requires high-quality test procedures that rely on the use of test features that increases the controllability and the observability of inner points of the circuit. Unfortunately, this is harmful from a security point of view, and therefore the access to these test features must be protected from unauthorized users. Another harm is related to the possibility for an untrusted manufacturer to do malicious alterations to the design (for instance to bypass or to disable the security fence of the system). Nowadays, many steps of the production cycle of a circuit are outsourced. For economic reasons, the manufacturing process is often carried out by foundries located in foreign countries. The threat brought by so-called Hardware Trojan Horses, which was long considered theoretical, begins to materialize.
A second issue is the hazard of faults that can appear during the circuit’s lifetime and that may affect the circuit behavior by way of soft errors or deliberate manipulations, called Fault Attacks. They can be based on the intentional modification of the circuit’s environment (e.g., applying extreme temperature, exposing the IC to radiation, X-rays, ultra-violet or visible light, or tampering with clock frequency) in such a way that the function implemented by the device generates an erroneous result. The attacker can discover secret information by comparing the erroneous result with the correct one. In-the-field detection of any failing behavior is therefore of prime interest for taking further action, such as discontinuing operation or triggering an alarm. In addition, today’s smart cards use 90nm technology and according to the various suppliers of chip, 65nm technology will be effective on the horizon 2013-2014. Since the energy required to force a transistor to switch is reduced for these new technologies, next-generation secure systems will become even more sensitive to various classes of fault attacks.
Based on these considerations, within the group I work with, we have proposed new methods, architectures and tools to solve the following problems:
• Test of secure devices: unfortunately, classical techniques for digital circuit testing cannot be easily used in this context. Indeed, classical testing solutions are based on the use of Design-For-Testability techniques that add hardware components to the circuit, aiming to provide full controllability and observability of internal states. Because crypto‐ processors and others cores in a secure system must pass through high‐quality test procedures to ensure that data are correctly processed, testing of crypto chips faces a dilemma. In fact design‐for‐testability schemes want to provide high controllability and observability of the device while security wants minimal controllability and observability in order to hide the secret. We have therefore proposed, form one side, the use of enhanced scan-based test techniques that exploit compaction schemes to reduce the observability of internal information while preserving the high level of testability. From the other side, we have proposed the use of Built-In Self-Test for such devices in order to avoid scan chain based test.
• Reliability of secure devices: we proposed an on-line self-test architecture for hardware implementation of the Advanced Encryption Standard (AES). The solution exploits the inherent spatial replications of a parallel architecture for implementing functional redundancy at low cost.
• Fault Attacks: one of the most powerful types of attack for secure devices is based on the intentional injection of faults (for instance by using a laser beam) into the system while an encryption occurs. By comparing the outputs of the circuits with and without the injection of the fault, it is possible to identify the secret key. To face this problem we have analyzed how to use error detection and correction codes as counter measure against this type of attack, and we have proposed a new code-based architecture. Moreover, we have proposed a bulk built-in current-sensor that allows detecting the presence of undesired current in the substrate of the CMOS device.
• Fault simulation: to evaluate the effectiveness of countermeasures against fault attacks, we developed an open source fault simulator able to perform fault simulation for the most classical fault models as well as user-defined electrical level fault models, to accurately model the effect of laser injections on CMOS circuits.
• Side-Channel attacks: they exploit physical data-related information leaking from the device (e.g. current consumption or electro-magnetic emission). One of the most intensively studied attacks is the Differential Power Analysis (DPA) that relies on the observation of the chip power fluctuations during data processing. I studied this type of attack in order to evaluate the influence of the countermeasures against fault attack on the power consumption of the device. Indeed, the introduction of countermeasures for one type of attack could lead to the insertion of some circuitry whose power consumption is related to the secret key, thus allowing another type of attack more easily. We have developed a flexible integrated simulation-based environment that allows validating a digital circuit when the device is attacked by means of this attack. All architectures we designed have been validated through this tool. Moreover, we developed a methodology that allows to drastically reduce the time required to validate countermeasures against this type of attack.
TSV- based 3D Stacked Integrated Circuits Test
The stacking process of integrated circuits using TSVs (Through Silicon Via) is a promising technology that keeps the development of the integration more than Moore’s law, where TSVs enable to tightly integrate various dies in a 3D fashion. Nevertheless, 3D integrated circuits present many test challenges including the test at different levels of the 3D fabrication process: pre-, mid-, and post- bond tests. Pre-bond test targets the individual dies at wafer level, by testing not only classical logic (digital logic, IOs, RAM, etc) but also unbounded TSVs. Mid-bond test targets the test of partially assembled 3D stacks, whereas finally post-bond test targets the final circuit.
The activities carried out within this topic cover 2 main issues:
• Pre-bond test of TSVs: the electrical model of a TSV buried within the substrate of a CMOS circuit is a capacitance connected to ground (when the substrate is connected to ground). The main assumption is that a defect may affect the value of that capacitance. By measuring the variation of the capacitance’s value it is possible to check whether the TSV is correctly fabricated or not. We have proposed a method to measure the value of the capacitance based on the charge/ discharge delay of the RC network containing the TSV.
• Test infrastructures for 3D stacked Integrated Circuits: testing a die before stacking to another die introduces the problem of a dynamic test infrastructure, where test data must be routed to a specific die based on the reached fabrication step. New solutions are proposed in literature that allow reconfiguring the test paths within the circuit, based on on-the-fly requirements. We have started working on an extension of the IEEE P1687 test standard that makes use of an automatic die-detection based on pull-up resistors.
Memory and Microprocessor Test and Reliability
Thanks to device shrinking and miniaturization of fabrication technology, performances of microprocessors and of memories have grown of more than 5 magnitude order in the last 30 years. With this technology trend, it is necessary to face new problems and challenges, such as reliability, transient errors, variability and aging.
In the last five years I’ve worked in cooperation with the Testgroup of Politecnico di Torino (Italy) to propose a new method to on-line validate the correctness of the program execution of a microprocessor. The main idea is to monitor a small set of control signals of the processors in order to identify incorrect activation sequences. This approach can detect both permanent and transient errors of the internal logic of the processor.
Concerning the test of memories, we have proposed a new approach to automatically generate test programs starting from a functional description of the possible faults in the memory.
Moreover, we proposed a new methodology, based on microprocessor error probability profiling, that aims at estimating fault injection results without the need of a typical fault injection setup. The proposed methodology is based on two main ideas: a one-time fault-injection analysis of the microprocessor architecture to characterize the probability of successful execution of each of its instructions in presence of a soft-error, and a static and very fast analysis of the control and data flow of the target software application to compute its probability of success.
Loading...