Common Weakness Enumeration

Example 1

The following SystemVerilog code is a crypto module that takes input data and encrypts it by processing the data through multiple encryption rounds. Note: this example is derived from [REF-1469].

In line 50 above, data_state_q is assigned to data_o. Since data_state_q contains intermediate state/results, this allows an attacker to obtain these results through data_o.

In line 50 of the fixed logic below, while "data_state_q" does not contain the final result, a "sanitizing" mechanism drives a safe default value (i.e., 0) to "data_o" instead of the value of "data_state_q". In doing so, the mechanism prevents the exposure of intermediate state/results which could be used to break soundness of the cryptographic operation being performed. A real-world example of this weakness and mitigation can be seen in a pull request that was submitted to the OpenTitan Github repository [REF-1469].

Many commodity processors have Instruction Set Architecture (ISA) features that protect software components from one another. These features can include memory segmentation, virtual memory, privilege rings, trusted execution environments, and virtual machines, among others. For example, virtual memory provides each process with its own address space, which prevents processes from accessing each other's private data. Many of these features can be used to form hardware-enforced security boundaries between software components.

When separate software components (for example, two processes) share microarchitectural predictor state across a hardware boundary, code in one component may be able to influence microarchitectural predictor behavior in another component. If the predictor can cause transient execution, the shared predictor state may allow an attacker to influence transient execution in a victim, and in a manner that could allow the attacker to infer private data from the victim by monitoring observable discrepancies (CWE-203) in a covert channel [REF-1400].

Predictor state may be shared when the processor transitions from one component to another (for example, when a process makes a system call to enter the kernel). Many commodity processors have features which prevent microarchitectural predictions that occur before a boundary from influencing predictions that occur after the boundary.

Predictor state may also be shared between hardware threads, for example, sibling hardware threads on a processor that supports simultaneous multithreading (SMT). This sharing may be benign if the hardware threads are simultaneously executing in the same software component, or it could expose a weakness if one sibling is a malicious software component, and the other sibling is a victim software component. Processors that share microarchitectural predictors between hardware threads may have features which prevent microarchitectural predictions that occur on one hardware thread from influencing predictions that occur on another hardware thread.

Features that restrict predictor state sharing across transitions or between hardware threads may be always-on, on by default, or may require opt-in from software.

Example 1

Branch Target Injection (BTI) is a vulnerability that can allow an SMT hardware thread to maliciously train the indirect branch predictor state that is shared with its sibling hardware thread. A cross-thread BTI attack requires the attacker to find a vulnerable code sequence within the victim software. For example, the authors of [REF-1415] identified the following code sequence in the Windows library ntdll.dll:

To successfully exploit this code sequence to disclose the victim's private data, the attacker must also be able to find an indirect branch site within the victim, where the attacker controls the values in edi and ebx, and the attacker knows the value in edx as shown above at the indirect branch site.

A proof-of-concept cross-thread BTI attack might proceed as follows:

1. The attacker thread and victim thread must be co-scheduled on the same physical processor core.
2. The attacker thread must train the shared branch predictor so that when the victim thread reaches indirect_branch_site, the jmp instruction will be predicted to target example_code_sequence instead of the correct architectural target. The training procedure may vary by processor, and the attacker may need to reverse-engineer the branch predictor to identify a suitable training algorithm.
3. This step assumes that the attacker can control some values in the victim program, specifically the values in edi and ebx at indirect_branch_site. When the victim reaches indirect_branch_site the processor will (mis)predict example_code_sequence as the target and (transiently) execute the adc instructions. If the attacker chooses ebx so that `ebx = m
   • 0x13BE13BD - edx, then the first adc will load 32 bits from address m in the victim's address space and add *m (the data loaded from) to the attacker-controlled base address in edi. The second adc instruction accesses a location in memory whose address corresponds to *m`.
4. The adversary uses a covert channel analysis technique such as Flush+Reload ([REF-1416]) to infer the value of the victim's private data *m.

Example 2

BTI can also allow software in one execution context to maliciously train branch predictor entries that can be used in another context. For example, on some processors user-mode software may be able to train predictor entries that can also be used after transitioning into kernel mode, such as after invoking a system call. This vulnerability does not necessarily require SMT and may instead be performed in synchronous steps, though it does require the attacker to find an exploitable code sequence in the victim's code, for example, in the kernel.

Many commodity processors have Instruction Set Architecture (ISA) features that protect software components from one another. These features can include memory segmentation, virtual memory, privilege rings, trusted execution environments, and virtual machines, among others. For example, virtual memory provides each process with its own address space, which prevents processes from accessing each other's private data. Many of these features can be used to form hardware-enforced security boundaries between software components.

Many commodity processors also share microarchitectural resources that cache (temporarily store) data, which may be confidential. These resources may be shared across processor contexts, including across SMT threads, privilege rings, or others.

When transient operations allow access to ISA-protected data in a shared microarchitectural resource, this might violate users' expectations of the ISA feature that is bypassed. For example, if transient operations can access a victim's private data in a shared microarchitectural resource, then the operations' microarchitectural side effects may correspond to the accessed data. If an attacker can trigger these transient operations and observe their side effects through a covert channel [REF-1400], then the attacker may be able to infer the victim's private data. Private data could include sensitive program data, OS/VMM data, page table data (such as memory addresses), system configuration data (see Demonstrative Example 3), or any other data that the attacker does not have the required privileges to access.

Example 1

Some processors may perform access control checks in parallel with memory read/write operations. For example, when a user-mode program attempts to read data from memory, the processor may also need to check whether the memory address is mapped into user space or kernel space. If the processor performs the access concurrently with the check, then the access may be able to transiently read kernel data before the check completes. This race condition is demonstrated in the following code snippet from [REF-1408], with additional annotations:

Vulnerable processors may return kernel data from a shared microarchitectural resource in line 4, for example, from the processor's L1 data cache. Since this vulnerability involves a race condition, the mov in line 4 may not always return kernel data (that is, whenever the check "wins" the race), in which case this demonstration code re-attempts the access in line 6. The accessed data is multiplied by 4KB, a common page size, to make it easier to observe via a cache covert channel after the transmission in line 7. The use of cache covert channels to observe the side effects of transient execution has been described in [REF-1408].

Example 2

Many commodity processors share microarchitectural fill buffers between sibling hardware threads on simultaneous multithreaded (SMT) processors. Fill buffers can serve as temporary storage for data that passes to and from the processor's caches. Microarchitectural Fill Buffer Data Sampling (MFBDS) is a vulnerability that can allow a hardware thread to access its sibling's private data in a shared fill buffer. The access may be prohibited by the processor's ISA, but MFBDS can allow the access to occur during transient execution, in particular during a faulting operation or an operation that triggers a microcode assist.

More information on MFBDS can be found in [REF-1405] and [REF-1409].

Example 3

Some processors may allow access to system registers (for example, system coprocessor registers or model-specific registers) during transient execution. This scenario is depicted in the code snippet below. Under ordinary operating circumstances, code in exception level 0 (EL0) is not permitted to access registers that are restricted to EL1, such as TTBR0_EL1. However, on some processors an earlier mis-prediction can cause the MRS instruction to transiently read the value in an EL1 register. In this example, a conditional branch (line 2) can be mis-predicted as "not taken" while waiting for a slow load (line 1). This allows MRS (line 3) to transiently read the value in the TTBR0_EL1 register. The subsequent memory access (line 6) can allow the restricted register's value to become observable, for example, over a cache covert channel.

Code snippet is from [REF-1410]. See also [REF-1411].

Device configuration controls are commonly programmed after a device power reset by a trusted firmware or software module (e.g., BIOS/bootloader) and then locked from any further modification. This is commonly implemented using a trusted lock bit, which when set, disables writes to a protected set of registers or address regions. The lock protection is intended to prevent modification of certain system configuration (e.g., memory/memory protection unit configuration). If debug features supported by hardware or internal modes/system states are supported in the hardware design, modification of the lock protection may be allowed allowing access and modification of configuration information.

Example 1

For example, consider the example Locked_override_register example. This register module supports a lock mode that blocks any writes after lock is set to 1.
However, it also allows override of the lock protection when scan_mode or debug_unlocked modes are active.

If either the scan_mode or the debug_unlocked modes can be triggered by software, then the lock protection may be bypassed.

Example 2

The following example code [REF-1375] is taken from the register lock security peripheral of the HACK@DAC'21 buggy OpenPiton SoC. It demonstrates how to lock read or write access to security-critical hardware registers (e.g., crypto keys, system integrity code, etc.). The configuration to lock all the sensitive registers in the SoC is managed through the reglk_mem registers. These reglk_mem registers are reset when the hardware powers up and configured during boot up. Malicious users, even with kernel-level software privilege, do not get access to the sensitive contents that are locked down. Hence, the security of the entire system can potentially be compromised if the register lock configurations are corrupted or if the register locks are disabled.

The example code [REF-1375] illustrates an instance of a vulnerable implementation of register locks in the SoC. In this flawed implementation [REF-1375], the reglk_mem registers are also being reset when the system enters debug mode (indicated by the jtag_unlock signal). Consequently, users can simply put the processor in debug mode to access sensitive contents that are supposed to be protected by the register lock feature.

This can be mitigated by excluding debug mode signals from the reset logic of security-critical register locks as demonstrated in the following code snippet [REF-1376].

Software commonly accesses peripherals in a System-on-Chip (SoC) or other device through a memory-mapped register interface. Malicious software could tamper with any security-critical hardware data that is accessible directly or indirectly through the register interface, which could lead to a loss of confidentiality and integrity.

Example 1

The register interface provides software access to hardware functionality. This functionality is an attack surface. This attack surface may be used to run untrusted code on the system through the register interface. As an example, cryptographic accelerators require a mechanism for software to select modes of operation and to provide plaintext or ciphertext data to be encrypted or decrypted as well as other functions. This functionality is commonly provided through registers.

Example 2

The example code is taken from the Control/Status Register (CSR) module inside the processor core of the HACK@DAC'19 buggy CVA6 SoC [REF-1340]. In RISC-V ISA [REF-1341], the CSR file contains different sets of registers with different privilege levels, e.g., user mode (U), supervisor mode (S), hypervisor mode (H), machine mode (M), and debug mode (D), with different read-write policies, read-only (RO) and read-write (RW). For example, machine mode, which is the highest privilege mode in a RISC-V system, registers should not be accessible in user, supervisor, or hypervisor modes.

The vulnerable example code allows the machine exception program counter (MEPC) register to be accessed from a user mode program by excluding the MEPC from the access control check. MEPC as per the RISC-V specification can be only written or read by machine mode code. Thus, the attacker in the user mode can run code in machine mode privilege (privilege escalation).

To mitigate the issue, fix the privilege check so that it throws an Illegal Instruction Exception for user mode accesses to the MEPC register. [REF-1345]

Isolated memory regions and access control (read/write) policies are used by hardware to protect privileged software. Software components are often allowed to change or remap memory region definitions in order to enable flexible and dynamically changeable memory management by system software.

If a software component running at lower privilege can program a memory address region to overlap with other memory regions used by software running at higher privilege, privilege escalation may be available to attackers. The memory protection unit (MPU) logic can incorrectly handle such an address overlap and allow the lower-privilege software to read or write into the protected memory region, resulting in privilege escalation attack. An address overlap weakness can also be used to launch a denial of service attack on the higher-privilege software memory regions.

Example 1

For example, consider a design with a 16-bit address that has two software privilege levels: Privileged_SW and Non_privileged_SW. To isolate the system memory regions accessible by these two privilege levels, the design supports three memory regions: Region_0, Region_1, and Region_2.

Each region is defined by two 32 bit registers: its range and its access policy.

• Address_range[15:0]: specifies the Base address of the region
• Address_range[31:16]: specifies the size of the region
• Access_policy[31:0]: specifies what types of software can access a region and which actions are allowed

Certain bits of the access policy are defined symbolically as follows:

• Access_policy.read_np: if set to one, allows reads from Non_privileged_SW
• Access_policy.write_np: if set to one, allows writes from Non_privileged_SW
• Access_policy.execute_np: if set to one, allows code execution by Non_privileged_SW
• Access_policy.read_p: if set to one, allows reads from Privileged_SW
• Access_policy.write_p: if set to one, allows writes from Privileged_SW
• Access_policy.execute_p: if set to one, allows code execution by Privileged_SW

For any requests from software, an address-protection filter checks the address range and access policies for each of the three regions, and only allows software access if all three filters allow access.

Consider the following goals for access control as intended by the designer:

• Region_0 & Region_1: registers are programmable by Privileged_SW
• Region_2: registers are programmable by Non_privileged_SW

The intention is that Non_privileged_SW cannot modify memory region and policies defined by Privileged_SW in Region_0 and Region_1. Thus, it cannot read or write the memory regions that Privileged_SW is using.

This design could be improved in several ways.

Example 2

The example code below is taken from the IOMMU controller module of the HACK@DAC'19 buggy CVA6 SoC [REF-1338]. The static memory map is composed of a set of Memory-Mapped Input/Output (MMIO) regions covering different IP agents within the SoC. Each region is defined by two 64-bit variables representing the base address and size of the memory region (XXXBase and XXXLength).

In this example, we have 12 IP agents, and only 4 of them are called out for illustration purposes in the code snippets. Access to the AES IP MMIO region is considered privileged as it provides access to AES secret key, internal states, or decrypted data.

The vulnerable code allows the overlap between the protected MMIO region of the AES peripheral and the unprotected UART MMIO region. As a result, unprivileged users can access the protected region of the AES IP. In the given vulnerable example UART MMIO region starts at address 64'h1000_0000 and ends at address 64'h1011_1000 (UARTBase is 64'h1000_0000, and the size of the region is provided by the UARTLength of 64'h0011_1000).

On the other hand, the AES MMIO region starts at address 64'h1010_0000 and ends at address 64'h1010_1000, which implies an overlap between the two peripherals' memory regions. Thus, any user with access to the UART can read or write the AES MMIO region, e.g., the AES secret key.

To mitigate this issue, remove the overlapping address regions by decreasing the size of the UART memory region or adjusting memory bases for all the remaining peripherals. [REF-1339]

A System-On-a-Chip (SoC) has a lot of functionality, but it may have a limited number of pins or pads. A pin can only perform one function at a time. However, it can be configured to perform multiple different functions. This technique is called pin multiplexing. Similarly, several resources on the chip may be shared to multiplex and support different features or functions. When such resources are shared between trusted and untrusted agents, untrusted agents may be able to access the assets intended to be accessed only by the trusted agents.

Example 1

Consider the following SoC design. The Hardware Root of Trust (HRoT) local SRAM is memory mapped in the core{0-N} address space. The HRoT allows or disallows access to private memory ranges, thus allowing the sram to function as a mailbox for communication between untrusted and trusted HRoT partitions.

We assume that the threat is from malicious software in the untrusted domain. We assume this software has access to the core{0-N} memory map and can be running at any privilege level on the untrusted cores. The capability of this threat in this example is communication to and from the mailbox region of SRAM modulated by the hrot_iface. To address this threat, information must not enter or exit the shared region of SRAM through hrot_iface when in secure or privileged mode.

In integrated circuits and hardware intellectual property (IP) cores, device configuration controls are commonly programmed after a device power reset by a trusted firmware or software module (e.g., BIOS/bootloader) and then locked from any further modification.

This behavior is commonly implemented using a trusted lock bit. When set, the lock bit disables writes to a protected set of registers or address regions. Design or coding errors in the implementation of the lock bit protection feature may allow the lock bit to be modified or cleared by software after it has been set. Attackers might be able to unlock the system and features that the bit is intended to protect.

Example 1

Consider the example design below for a digital thermal sensor that detects overheating of the silicon and triggers system shutdown. The system critical temperature limit (CRITICAL_TEMP_LIMIT) and thermal sensor calibration (TEMP_SENSOR_CALIB) data have to be programmed by firmware, and then the register needs to be locked (TEMP_SENSOR_LOCK).

In this example, note that if the system heats to critical temperature, the response of the system is controlled by the TEMP_HW_SHUTDOWN bit [1], which is not lockable. Thus, the intended security property of the critical temperature sensor cannot be fully protected, since software can misconfigure the TEMP_HW_SHUTDOWN register even after the lock bit is set to disable the shutdown response.

Example 2

The following example code is a snippet from the register locks inside the buggy OpenPiton SoC of HACK@DAC'21 [REF-1350]. Register locks help prevent SoC peripherals' registers from malicious use of resources. The registers that can potentially leak secret data are locked by register locks.

In the vulnerable code, the reglk_mem is used for locking information. If one of its bits toggle to 1, the corresponding peripheral's registers will be locked. In the context of the HACK@DAC System-on-Chip (SoC), it is pertinent to note the existence of two distinct categories of reset signals.

First, there is a global reset signal denoted as "rst_ni," which possesses the capability to simultaneously reset all peripherals to their respective initial states.

Second, we have peripheral-specific reset signals, such as "rst_9," which exclusively reset individual peripherals back to their initial states. The administration of these reset signals is the responsibility of the reset controller module.

In the buggy SoC architecture during HACK@DAC'21, a critical issue arises within the reset controller module. Specifically, the reset controller can inadvertently transmit a peripheral reset signal to the register lock within the user privilege domain.

This unintentional action can result in the reset of the register locks, potentially exposing private data from all other peripherals, rendering them accessible and readable.

To mitigate the issue, remove the extra reset signal rst_9 from the register lock if condition. [REF-1351]

A device might support features such as secure boot which are supplemented with hardware and firmware support. This involves establishing a chain of trust, starting with an immutable root of trust by checking the signature of the next stage (culminating with the OS and runtime software) against a golden value before transferring control. The intermediate stages typically set up the system in a secure state by configuring several access control settings. Similarly, security logic for exercising a debug or testing interface may be implemented in hardware, firmware, or both. A device needs to guard against fault attacks such as voltage glitches and clock glitches that an attacker may employ in an attempt to compromise the system.

Example 1

Below is a representative snippet of C code that is part of the secure-boot flow. A signature of the runtime-firmware image is calculated and compared against a golden value. If the signatures match, the bootloader loads runtime firmware. If there is no match, an error halt occurs. If the underlying hardware executing this code does not contain any circuitry or sensors to detect voltage or clock glitches, an attacker might launch a fault-injection attack right when the signature check is happening (at the location marked with the comment), causing a bypass of the signature-checking process.

After bypassing secure boot, an attacker can gain access to system assets to which the attacker should not have access.

• Power
• Clock

An adversary could monitor and measure physical phenomena to detect patterns and make inferences, even if it is not possible to extract the information in the digital domain.

Physical side channels have been well-studied for decades in the context of breaking implementations of cryptographic algorithms or other attacks against security features. These side channels may be easily observed by an adversary with physical access to the device, or using a tool that is in close proximity. If the adversary can monitor hardware operation and correlate its data processing with power, EME, and acoustic measurements, the adversary might be able to recover of secret keys and data.

Example 1

Consider a device that checks a passcode to unlock the screen.

PIN numbers used to unlock a cell phone should not exhibit any characteristics about themselves. This creates a side channel. An attacker could monitor the pulses using an oscilloscope or other method. Once the first character is correctly guessed (based on the oscilloscope readings), they can then move to the next character, which is much more efficient than the brute force method of guessing every possible sequence of characters.

Example 2

Consider the device vulnerability CVE-2021-3011, which affects certain microcontrollers [REF-1221]. The Google Titan Security Key is used for two-factor authentication using cryptographic algorithms. The device uses an internal secret key for this purpose and exchanges information based on this key for the authentication. If this internal secret key and the encryption algorithm were known to an adversary, the key function could be duplicated, allowing the adversary to masquerade as the legitimate user.

Example 3

The code snippet provided here is part of the modular exponentiation module found in the HACK@DAC'21 Openpiton System-on-Chip (SoC), specifically within the RSA peripheral [REF-1368]. Modular exponentiation, denoted as "a^b mod n," is a crucial operation in the RSA public/private key encryption. In RSA encryption, where 'c' represents ciphertext, 'm' stands for a message, and 'd' corresponds to the private key, the decryption process is carried out using this modular exponentiation as follows: m = c^d mod n, where 'n' is the result of multiplying two large prime numbers.

The vulnerable code shows a buggy implementation of binary exponentiation where it updates the result register (result_reg) only when the corresponding exponent bit (exponent_reg[0]) is set to 1. However, when this exponent bit is 0, the output register is not updated. It's important to note that this implementation introduces a physical power side-channel vulnerability within the RSA core. This vulnerability could expose the private exponent to a determined physical attacker. Such exposure of the private exponent could lead to a complete compromise of the private key.

To address mitigation requirements, the developer can develop the module by minimizing dependency on conditions, particularly those reliant on secret keys. In situations where branching is unavoidable, developers can implement masking mechanisms to obfuscate the power consumption patterns exhibited by the module (see good code example). Additionally, certain algorithms, such as the Karatsuba algorithm, can be implemented as illustrative examples of side-channel resistant algorithms, as they necessitate only a limited number of branch conditions [REF-1369].

• Power