HOD HASHARON, Israel and DETROIT–Sept. 29, 2016–Karamba Security today announced Autonomous Security for connected and autonomous vehicles, which empowers their electronic control units (ECUs) to protect themselves from hackers. Autonomous Security, a significant extension to the company’s Carwall ECU security platform, enables automotive technology providers to achieve the goals set out in the U.S. Department of Transportation’s guidelines for the safe deployment of autonomous cars.

Separately, Karamba Security also announced a new strategic investment round led by Fontinalis Partners, a firm solely focused on investing in and scaling technology companies that are advancing next-generation mobility solutions.

New research from Navigant Research projects that sales of connected and fully autonomous vehicles will grow from 14 million annually in 2020, about 15 percent of annual car sales globally, to nearly 72 million in 2025, accounting for nearly 70 percent of light duty vehicles. By the mid-2020s, they forecast more than 245 million vehicles with level 2 connectivity or higher will be on the road.

In the report, “Autonomous Automotive Cybersecurity,” Senior Research Analyst Sam Abuelsamid concludes solving cybersecurity problems, however, is critical for achieving this forecast, and one key risk he singles out is the probability of having false positives. With today’s network-based approaches, approximately two of every 100 commands suspected as malicious on the car’s CAN bus network will be mistakenly blocked. “The fatality risk of such frequent blocking of valid operations is something the automotive industry cannot tolerate,” he said.

Karamba Security’s automated ECU technology eliminates this risk by providing zero false positives.

Stephan A. Tarnutzer, vice president, Electronics at FEV North America, Inc., a Tier-1 global automotive supplier and internationally recognized powertrain and vehicle engineering company, agrees with Abuelsamid that preventing false positives is a critical point.

“Experiencing false positives or detecting hacks on the CAN bus, only after the fact, are unacceptable risks in vehicle engineering,” said Tarnutzer. “FEV North America works with Karamba Security because we’ve seen the benefit of its Autonomous Security technology in securing our own ECUs. Karamba’s technology doesn’t require any developer resources to install or generate the security policy, and its CPU footprint is negligible. We are working with Karamba to integrate Carwall into our reference platform, which will allow us to present our automotive customers with a secure system out of the box.”

How Carwall Autonomous Security works

Cyberattacks can only infiltrate a car by compromising the externally-connected ECUs controlling infotainment, navigation and OBDII telematics dongles, for example.

Karamba Security’s Autonomous Security technology allows any car’s ECU to protect itself from this threat by automatically locking it down to the ECU’s factory settings. The ECU then blocks operations that aren’t part of its factory settings, with a negligible performance impact, which prevents hackers from accessing the car’s safety systems and commandeering them.

This deterministic decision is made locally on the ECU. Autonomous Security doesn’t require the ECU be connected to protect itself, nor does it need anti-malware updates.

Today, Karamba Security also unveiled a new capability, in-memory protection, as part of its Autonomous Security suite. With in-memory protection, the ECU autonomously blocks memory-based attacks such as buffer overrun and return oriented programming (ROP).

In-memory Autonomous Security blocks common in-memory attacks, such as the Tesla hack demonstrated last week. With its recently announced security upgrade, Tesla has stated that it tried to make it harder for hackers to reprogram other ECUs once they’ve compromised the externally connected ECU. In effect, Tesla’s approach accepts that hackers will penetrate the car’s ECU and then tries to minimize the damage. In sharp contrast, Karamba Security in-memory protection blocks such hacks altogether. Specifically, the attack demonstrated by the researchers would have failed in Teslas or any vehicle protected by Carwall.