edfas.org ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 4 28 active layer, where materials change their properties. A ReRAM cell uses bistable resistance state oxide materials like NiO, ZrO2, and HfO2. Ferroelectric HfO2 has polarization properties, but no ferromagnetic properties, existing inside the device. With the MOCI technique, polarized light is reflected from the magnetic materials. Therefore, ReRAM is not vulnerable to theMOCI technique if HfO2 has been used. However, if NiOmaterial (which has ferromagnetic properties) is used as data storage layer in ReRAM, data can be read out using MOCI techniques. Therefore, when insulating layer NiO exists, ReRAM may be potentially vulnerable to MOCI. Furthermore, at different bias conditions, filament is formed through insulating oxide in the cell, and NIR photon emission can be used to determine the physical locationof the filaments fromReRAMcells. This alsomakes ReRAM vulnerable to photon emission microscopy. Energy dispersive spectroscopy (EDX) may potentially be used to decode the bit information stored in the metal oxide layer, as EDX technique is used for elemental analysis. The metal oxide layer in LRS formation contains a conductive filament made up of oxygen vacancies. On the other hand, due to filament rapture in HRS, oxygen migrates on the top surface. The broken conductive filament ensures oxygen existence can be detected by EDX analysis, ensuring a high resistance (“on”) state. EDX cannot detect oxygen when it no longer exists in conductive filament, ensuring a low resistance (“off”) state. Therefore, EDX can potentially be used to read bit information stored in themetal oxide layer by determining the physical location of the filaments. The conductivity of polymer materials is a function of resistance. Given how they function, STEM and CAFM imaging can reveal thebit informationof polymermemory cells bymeasuring current. Since nanowires are ferromagnetic and the change in thepolarizationoccurs innanowire regions, racetrack memory may be vulnerable to MOCI and QDM imaging techniques. Carbon memory (NRAM) and mott memory are still undergoing research and development in their early stage. However, in mott memory, the phase change of materials may be determined by CAFM. COUNTERMEASURES AND RESEARCH OPPORTUNITIES Researchers have proposed techniques to shield against physical attacks that make use of imaging modalities. These modalities are constantly evolving, so there is still a prominent concern of the threat against emergingNVM technologies. Unfortunately, many of the proposed works do not take into consideration the growing complexity of the emerging technologies and fail to address the more advanced FA tools that have entered the arena in recent years. This section will discuss some of the potential countermeasures against physical attacks from literature and from our research. The discussion can be separated into three solution categories: device level, design level, and package level. Table 2 shows the existence of possible solutions at different levels of design. At device-level, nanopyramid structures can be inserted for further protection. During the fabrication process, susceptible areas of the die are selected for this enhancement. Randomly distributed silicon pyramids are then inserted, which interfere with light absorption, scattering, and reflection elements that are crucial to optical-based attacks. By tampering with hypothetical attacks in such a manner, malicious data extraction is nullified.[14] Additionally, device-based methods may involve protective shields, a concept compatible with modern IC packaging techniques and compact systems like IoT devices.[15] In actuality, the shielding effect of all materials depends on the frequencyof the radiation. Some materials are best suited formicrowaves, while others are better suited for lower frequencies.[16] Thismaterials-based approach can be used to prevent sMIM and SCM based attacks. Furthermore, the security properties of novel materials can be exploited as tamper-proof systems. For example, antiferromagnetic materials and superconductors are known to be invulnerable to or expel magnetic fields, respectively.[11] Therefore, an antiferromagnetic material-based memory system could be capable of preventing electromagnetic probing attacks. A design-based countermeasure can also be used to detect backside polishing attempts. Sample preparation is a common step used before executing an attack on a target, which can disable/tamper a sensor circuit. A backside polishing detector (BPD) uses through silicon vias (TSVs) to monitor parasitic capacitance. Due to polishing Table 2 Types of physical attacks and possible level of solutions Types of attacks Solutions Device level Design level Package level SEM/STEM Yes Yes MOCI/QDM Yes sMIM/SCM Yes Yes NIR photon emission Yes Yes
RkJQdWJsaXNoZXIy MTMyMzg5NA==