The goal of tamper mechanisms is to prevent any attempt by an attacker to perform an unauthorized physical or electronic action against the device. Tamper mechanisms are divided into four groups:
Tamper mechanisms are most effectively used in layers to prevent access to any critical components. They are the primary facet of physical security for embedded systems and must be properly implemented to be successful. From the designer's perspective, the costs of a successful attack should outweigh the potential rewards.
Often, existing tamper mechanisms can only be discovered by attempted or complete disassembly of the target product. This may require an attacker to obtain more than one device in order to sacrifice one for the sole purpose of discovering such mechanisms. Once the mechanisms are noted, an adversary can form hypotheses about how to attack and bypass them.
Weingart's "Physical Security Devices for Computer Subsystems: A Survey of Attacks and Defenses"  describes physical tamper mechanism attacks and defenses ranging from cheap and easy to extremely costly and complex. It is a comprehensive guide to many (if not all) known attack types and provides lists of solutions to implement to protect against such attacks. Physical attacks include different types of probing (passive, active/injector, pico-probes, or energy); machining methods (manual material removal, mechanical, water, laser, chemical, or shaped charge); and electrical (radiation imprinting, temperature imprinting, high voltage imprinting, power supply fluctuations, clock glitching, circuit disruption, or electron beam and infrared laser read/write). Corresponding countermeasures for these attacks are described, including various types of physical barriers (hard barriers, single chip coatings, or insulator-based substrates); tamper evident solutions (brittle packages, crazed aluminum, polished packages, bleeding paint, or holographic tape); tamper detection sensors (voltage, probe, wire, printed circuit board, flex, stressed glass, piezo-electric, motion, ultrasonic, microwave, infrared, acceleration, radiation, flux, dosage, or temperature); and tamper response technologies (RAM power drop, RAM overwrite, or physical destruction).
Tamper resistance consists of using specialized materials to make tampering of a device or module difficult. This can include such features as hardened steel enclosures, locks, encapsulation, or security screws. Implementing tight airflow channels (that is, tightly packing the components and circuit boards within the enclosure) will increase the difficulty of optical probing of the product internals using fiber optics. A side benefit of many tamper resistant mechanisms is that they are often tamper evident, meaning that physical changes can be visually observed and it becomes obvious that the product has been tampered with.
If designing a housing that requires screws, or when retrofitting a design that is already using screws, consider implementing one-way screws that will offer additional tamper resistance. Although an adversary can likely drill through such screws, they raise the difficulty of attack over an industry-standard screwdriver or Torx driver bit. The Thomas Register Directory provides a large listing of security- and tamperproof-screw manufacturers and suppliers.
Sealing both sides of the housing together in a way that requires the destruction of the device in order to open it should be considered. Many plastics are sensitive to heat and melt at fairly low temperatures. Consider sealing the housing with high-temperature glue or ultrasonic welding to reduce tampering. If using high-temperature glue, choose one with a higher softening point than the plastic housing in order to increase visible tamper evidence. Serviceability may be an issue if the product is intended to be opened by authorized personnel. However, if a legitimate user can open the device, so can an adversary can.
An entire circuit board with resistant resin or epoxy compound can protect the circuitry. However, it is more common for such encapsulation to be done on only specific critical components. Conformal coatings and encapsulates are typically used to protect an assembled circuit board from moisture, fungus, dust, corrosion, or tampering. It can also reduce mechanical stress on components and protect them from thermal shock. Urethane provides a hard, durable coating that offers excellent abrasion and solvent resistance. It shrinks significantly during coating, however, which may stress components. Epoxies also offer excellent resistance to moisture and solvents. Usually consisting of a two-part thermosetting resin, the coating also shrinks during curing, leaving a hard, difficult-to-remove film. Conformal coatings are provided by a large number of manufacturers, including GE Silicones, Dow Corning, and MG Chemicals.
Chemicals such as methylene chloride, sulfuric acid, and fuming nitric acid can remove protective coatings, so be sure to evaluate that your chosen compound is suitable for your desired protection level. To protect against a chemical attack that removes the encapsulation, aluminum powder can be added to the compound. A solvent capable of dissolving the aluminum will corrode the underlying components or circuitry, rendering the device useless.
The goal of tamper evidence is to ensure that visible evidence is left behind when tampering occurs. Tamper evident mechanisms are a major deterrent for minimal risk takers (e.g., non-determined attackers). Hundreds of tamper evident materials and devices are available, mostly consisting of special seals and tapes to make it obvious that there has been physical tampering.
Tamper evidence features are only successful if a process is in place to check whether tampering has occurred or if a legitimate owner of the device notices a deformity. Generally speaking, if an adversary purchases a product with the specific intention of attacking it, tamper evident mechanisms by themselves will not prevent the attack.
Weingart's "Physical Security Devices for Computer Subsystems: A Survey of Attacks and Defenses"  provides dozens of potential tamper evident mechanisms to employ. Most (if not all) of the available tamper evident seals can be bypassed. In Johnston and Garcia's "Vulnerability Assessment of Security Seals,"  the authors show how 94 different security seals (including adhesive tape, plastic, wire loop, metal cable, metal ribbon, bolt type, secure container, passive fiber optic, and electronic) were defeated using low-cost tools and readily available supplies.
Holdtite manufactures Secure 42, superglue intended to provide evidence of tampering. Brittle plastics or enclosures that crack or shatter upon an attempted penetration may be suitable in certain environments. "Bleeding" paint, where paint of one color is mixed with tiny spheres of a contrasting color paint that rupture when the surface is scratched, is a novel solution.
Tamper detection mechanisms enable the hardware device to be aware of tampering and typically fall into one of three groups:
Switches such as microswitches, magnetic switches, mercury switches, and pressure contacts to detect the opening of a device, the breach of a physical security boundary, or the movement of a particular component.
Sensors such as temperature and radiation sensors to detect environmental changes, voltage and power sensors to detect glitch attacks, radiation sensors for X-rays (used for seeing what is inside of a sealed or encapsulated device) and ion beams (often used for advanced attacks to focus on specific electrical gates within an integrated circuit).
Circuitry such as flexible circuitry, nichrome wire, and fiber optics wrapped around critical circuitry or specific components on the board. These materials are used to detect a puncture, break, or attempted modification of the wrapper. For example, if the resistance of the nichrome wire changes or the light power traveling through the optical cable decreases, the system can assume there has been physical tampering.
Again, Weingart's "Physical Security Devices for Computer Subsystems: A Survey of Attacks and Defenses"  provides a comprehensive list of specific mechanisms that could be employed.
Tamper response mechanisms are the countermeasures taken upon the detection of tampering. Chaum's 1983 "Design Concepts for Tamper Responding Systems"  presents concepts for implementing sensors into tamper responsive systems.
Most often, the response consists of completely shutting down or disabling the device, or erasing critical portions of memory to prevent an attacker from accessing secret data. Physical destruction of a device using a small explosive charge may be an option for extremely secure devices, but is not practical for most (if any) consumer electronics. Response mechanisms may also be simpler, such as just logging the type of attack detected and the time it occurred, which can provide useful audit information and help with forensic analysis after an attack.
Simply erasing critical portions of memory (also known as "zeroizing") is usually not enough, however, as shown by Gutmann's "Secure Deletion of Data from Magnetic and Solid-State Memory"  and "Data Remanence in Semiconductor Devices,"  along with Skorobogatov's "Low Temperature Data Remanence in Static RAM."  Gutmann observes that "contrary to conventional wisdom, volatile semiconductor memory does not entirely lose its contents when power is removed. Both static (SRAM) and dynamic (DRAM) memory retains some information."
W.L. Gore's D3 electronic security enclosures are designed to protect the physical security boundary of a module and combine a number of tamper evidence and detection features. The sensor comes as a foldable sheet that is to be wrapped around the product. Conductive ink crisscrosses through the sheet with a maximum distance between traces of 200 to 300 microns (a pitch too small to be drilled through without detection). The electrical state of the sensor changes if the field is broken, which will trigger the product to enable its tamper respondent mechanisms. Gore claims that the device is transparent to X-rays (which may be used to determine the location of the sensor within the product) and that it has been tested against a wide range of reagents and solvents. The outer layer has an opaque resin coating, which conceals all surface details of the sensor and prevents an attacker from seeing any traces. This product also meets the requirements of FIPS 140 Level 4 Specification for Cryptographic Modules. 
Tamper response mechanisms are unlikely to trigger accidentally. Still, the legitimate user will need to understand the environmental and operational conditions and keep the device within those limits. Many tamper-responsive devices are designed and manufactured with the stipulation that they will never be openedlegitimately or not.