VaultIP Secures All Things
Inside Secure Offers a Range of Security IP for IoT Processors
By Loyd Case
Internet-connected refrigerators recently took some heat for allegedly being part of a botnet that sent out high volumes of spam. The fridge botnet turned out to be a false alarm, but the specter of compromised network-connected devices serving nefarious purposes no longer seems like the stuff of bad science fiction. One report noted the lack of security with Internet-connected baby-monitoring cameras, allowing intruders to use them to generate attacks on other network-connected devices.
Inside Secure’s VaultIP family gives processor designers a tool to secure Internet of Things (IoT) devices. The family consists of synthesizable intellectual property (IP) that customers can add to SoC designs, plus software tools based on existing cryptographic standards. The IP includes fixed-function cryptographic accelerators, secure lockboxes for key storage, secure policy-based key usage, and secure-boot capability. It ships in three flavors targeting different security features and available die areas. The light version, called the VaultIP-100, includes strong device authentication, secure key storage, and a true-random-number generator (TRNG). The company estimates this version consumes about 50,000 gates, or about 0.05mm2 in a 40nm technology.
The intermediate version, the VaultIP-120, adds secure communication with peer devices. It ships with a cryptographic accelerator that supports AES-256 encryption attached to a secure DMA for on-the-fly bulk data encryption. The VaultIP-120 requires about 100,000 gates, which translates to 0.10mm2 of die area in a 40nm technology.
Both versions are subsets of a larger hardware-IP offering called the VaultIP-130. Originally developed for mobile application processors, the -130 adds more cryptographic algorithms to its hardware accelerator, including 3DES and RSA. It also integrates with ARM’s TrustZone, bringing hardware cryptographic acceleration to TrustZone’s secure-execution environment.
Software tools include asset-policy management stored in secure firmware and managed by VaultIP’s controller. The controller is part of the hardware IP and consists of a simple sequencer and ALU; it’s not intended to run applications. Secure boot combines software libraries embedded in the processor ROM.
Customers licensing VaultIP get the Verilog RTL libraries, embedded firmware, drivers, the development kit, verification environment, test vectors, and middleware to connect the subsystem to the rest of the SoC. They can opt for additional personalization and provisioning, which the company provides as licensable tool kits or as a service.
Parsing IoT Threats
IoT devices address a wide range of use cases, some trivial and others not so trivial. A hacked smart lightbulb or toy may cause minor inconveniences, but a network of compromised medical devices could threaten the health or even life of patients. Hackers gain access through a variety of methods, including poorly implemented client security and man-in-the-middle attacks, as Figure 1 shows. Developers often focus on the added cost of implementing robust security, particularly when the price of a client IoT processor is under $1. But they should also focus on how a breach could damage their reputation and finances.
Figure 1. Network security threats. Hackers can gain network access at several different points. Developers, however, have a better understanding of security needs in the network and on the server side than at the IoT client.
Network complexity also gives hackers the opportunity to gain access. The Internet of Things consists of more than devices at the end points: as Figure 1 shows, the client device starts the data path, which passes through IoT gateways, communications networks, routers, and finally the cloud. Security breaches can happen anywhere along this path.
Client-side security frequently remains a work in progress, since IoT consists of many emerging platforms. The ecosystem of IoT developers typically includes many startups and smaller OEMs, often working on their first projects. Security must be easy to implement and built in, since these smaller developers lack the time and resources to build their own security. Integrating this function in the IoT processor enables developers to more easily implement it.
Modular, Compact Hardware Security
Large processor companies build security directly into their processors; ARM’s TrustZone and Qualcomm’s Secure Execution Environment (QSEE) are two examples. The former uses hardware virtualization to create a secure environment in ARM-based SoCs, obviating the need for a dedicated security processor. The latter comprises a suite of security solutions ranging from a trusted execution environment based on TrustZone to end-user authentication solutions.
TrustZone and QSEE lack specific IoT support. Semiconductor companies view purpose-built IoT processors as a substantial growth market. Given the potential volume, adding the right level of security for targeted applications seems appropriate. The VaultIP-130’s modular approach, shown in Figure 2, might be just the ticket, particularly for companies that can’t afford to develop their own security IP. ARM has recently announced TrustZone-M for its line of Cortex-M processors, often used in microcontrollers for IoT applications. VaultIP can complement TrustZone-M with a hardware trust anchor.
Figure 2. VaultIP-130 block diagram. Modules in this device include cryptographic accelerators, key storage, and ARM TrustZone-M support. Users can set up local or remote provisioning as needed.
Consumer applications, such as connected toys, often need less-robust security. These products run fixed code rather than multiple applications, with peer-to-peer connections and over-the-air updates being uncommon. They use tiny microcontrollers and typically transfer small amounts of data, making them ideal candidates for the VaultIP-100.
The VaultIP-100 contains enough cryptographic muscle to handle device authentication, which embeds in each chip a unique identifier secured via ECDSA (elliptic-curve digital-signature algorithm). ECDSA uses a smaller public key than traditional DSA, but it employs the same private-key size, building on more-complex algorithms than traditional RSA-style alternatives. ECDSA offers improved security strength compared with standard RSA algorithms while using significantly smaller keys. It also resists brute-force attacks better than RSA. Bitcoin, for example, employs ECDSA-based algorithms for encryption. In addition, the VaultIP-100 can perform secure authentication on the basis of symmetric algorithms thanks to its embedded SHA-256 crypto accelerator, as Figure 3 shows.
Figure 3. VaultIP-100 block diagram. Modules include hardware support for key protection and secure authentication. The embedded SHA-256 accelerator provides the cryptographic muscle.
The IP includes secure key storage in firmware sandboxed from the rest of the processor. It assumes the processor has some level of native cryptographic acceleration—enough to provide secure boot. The host CPU handles any symmetric cryptography for communicating with peripherals (such as sensors).
Adding Muscle With VaultIP-120/130
If the processor lacks the required cryptographic performance, Inside Secure recommends the VaultIP-120 or VaultIP-130. The 120 includes more layers of protection than the VaultIP-100, as Figure 4 shows. It adds secure communication, so peers can connect via the network with a high degree of trust. Moreover, the application can encrypt communication between peers to facilitate data security. Also included is secure storage for AES encryption keys. The 130 also performs software-integrity checks on signed code to ensure any executed code is valid; it also has a FIPS-certified TRNG for generating keys and offers legacy crypto like 3DES and SHA-1.
Figure 4. VaultIP-120 block diagram. This next step up from the VaultIP-100 adds secure communication. It has an on-board AES accelerator plus a high-speed DMA controller.
Inside Secure also builds in power management, so the accelerators and other functional units will sleep when they’re unneeded, consuming little power. The VaultIP-120 integrates a high-speed DMA controller to quickly transfer cryptographic and hash data. Both versions connect with application processors via standard Amba interfaces, such as AHB or AXI. Table 1 outlines the main features of Inside Secure’s three VaultIP products.
Table 1. Comparison of VaultIP versions. The VaultIP-100 enables device authentication. The 120 adds secure communication with fast DMA for moving cryptographic data between the accelerators and the host processor VaultIP-130 adds secure boot and additional cryptographic algorithms, as well as more levels of policy management. (Source: Inside Secure)
The company also provides services to facilitate proper setup of key storage and master keys during fabrication. It performs personalization, which inserts unique identifiers, on site during fabrication, testing, or packaging. Provisioning services can also be cloud based. Manufacturers can securely inject master keys and individual identifiers in the field or even locally in the chip, though the process still requires an external source for master keys.
The VaultIP-130 has achieved the rigorous FIPS 140-2 Level 2 certification, the first for a silicon-IP provider. This achievement allows chipmakers to apply for only an incremental certification rather than going through a full FIPS process, cutting time to market.
VaultIP can work with any existing CPU architecture to create a secure IoT processor. Figure 5 shows how a designer might integrate secure-boot capability into a SoC. The chip boots from secure on-chip ROM, which houses keys linked to that specific chip. It obtains a digitally signed and optionally encrypted version of the application from flash memory and validates it using VaultIP’s hardware integrity and authentication functions (ECDSA and SHA-256). The IP stores and uses the cryptographic keys, which are never exposed outside of this protected boundary. If VaultIP detects no errors, the application is moved to RAM, where it can run normally. Inside Secure provides these secure boot components as library elements called SafeZone.
Figure 5. Implementing secure boot. CPUs can link to the secure-boot IP from Inside Secure using a direct interconnect. The rose-colored blocks represent VaultIP technology.
The VaultIP-130 can also implement TrustZone and TrustZone-M capabilities (see MPR 11/16/15, “ARM Dons Thicker Armor”), providing hardware acceleration for ARM-based SoCs; Inside Secure participates in ARM’s TrustZone Ready program. TrustZone defines the capabilities of a trusted system, whereas Inside Secure provides the software stacks and IP to implement that system in an application processor. The technology also defines features beyond those that IoT processors need, such as content protection (DRM) and trusted execution environments (TEEs). These features may be required in set-top boxes and other devices that execute third-party apps or play protected audio and video files. But few IoT applications require this type of security, as they don’t perform such functions.
Who Needs Security?
As IoT enters the real world, concerns about security become more urgent. IoT application designers must consider the implications of security breaches and ensure their devices, however humble, offer enough protection to give customers peace of mind. A low-power, low-cost solution such as VaultIP makes implementing basic security features easier, both in the chip and in the application software.
Convincing chip designers to implement security is a challenge, however. No processor designer likes to add gates, which increase the overall die area and cost. Testing and validating security features can delay product shipments at a time when companies are racing to bring new ideas to market. Consumers today place little value on these features, allowing system vendors to ignore them.
But the cost of failing to implement security in today’s vulnerable devices may be much greater in the long term—and not just in dollars. Thieves could use a home’s own security cameras to case the property and then loot it when no one is there. A hacker who gains control of an automobile or medical device could cause potentially deadly consequences. When such serious breaches begin to happen, consumers will suddenly demand secure IoT devices. Companies that have already implemented robust security will then have the advantage.
Price and Availability
RTL for the VaultIP-130 is available now. The VaultIP-120 and VaultIP-100 are scheduled for RTL availability in 1Q16. We expect products in 4Q16. Inside Secure does not disclose licensee fees. For more information, access www.insidesecure.com/Products-Technologies/Silicon-IP/VaultIP.