#aixcc

At DEF CON, Team Atlanta’s AI system crushed the competition by automatically detecting and patching security bugs in massive codebases, signalling a leap forward in protecting critical software.

Source: The Record

DARPA AI cyber challenge

Among the most obvious advantages of artificial intelligence (AI) is its capacity to improve cybersecurity for businesses and global internet users. This is particularly true since malicious actors are still focussing on and taking advantage of vital software and systems. These dynamics will change as  artificial intelligence develops, and if used properly,  AI may help move the sector towards a more secure architecture for the digital age.

Experience also teaches us that, in order to get there, there must be strong public-private sector partnerships and innovative possibilities such as DARPA’s AI Cyber Challenge (AIxCC) Semifinal Competition event, which will take place from August 8–11 at DEF CON 32 in Las Vegas. Researchers in cybersecurity and AI get together for this two-year challenge to create new AI tools that will aid in the security of significant open-source projects.

This competition is the first iteration of the challenge that DARPA AIxCC unveiled at Black Hat last year. Following the White House’s voluntary AI commitments, which saw business and government unite to promote ethical methods in AI development and application, is the Semifinal Competition. Today, we’re going to discuss how Google is assisting rivals in the AIxCC:

Google  Cloud resources: Up to $1 million in Google Cloud credits will be awarded by Google to each qualifying AIxCC team, in an effort to position rivals for success. The credits can be redeemed by participants for Gemini and other qualified Google Cloud services. As a result, during the challenge, competitors will have access to Google Cloud AI and machine learning models, which they can utilise and expand upon for a variety of purposes. Google is urge participants to benefit from Google incentive schemes, such as the Google for firms Program, which reimburses up to $350,000 in Google Cloud expenses for AI firms.

Experts in cybersecurity: For years, Google has been in the forefront of utilising AI in security, whether it is for sophisticated threat detection or shielding Gmail users from phishing scams. Google excited to share knowledge with AIxCC by making security specialists available throughout the challenge, they have witnessed the promise of AI for security. Google security specialists will be available to offer guidance on infrastructure and projects as well as assist in creating the standards by which competitors will be judged. Specifically, Google discuss recommended approaches to AIxCC in Google tutorial on how AI may improve Google’s open source OSS-Fuzz technology.

Experience AIxCC at DEF CON: We’ll be showcasing Google’s AI technology next week at the AIxCC Semifinal Experience at DEF CON, complete with an AI-focused hacker demo and interactive demonstrations. In order to demonstrate AI technologies, which includes Google Security Operations and SecLM, Google will also have Chromebooks at the Google exhibit. Google security specialists will be available to guests to engage in technical conversations and impart information.

AI Cyber Challenge DARPA

At Black Hat USA 2023, DARPA invited top computer scientists,  AI experts, software engineers, and others to attend the AI Cyber Challenge (AIxCC). The two-year challenge encourages AI-cybersecurity innovation to create new tools.

Software powers everything in Google ever-connected society, from public utilities to financial institutions. Software increases productivity and makes life easier for people nowadays, but it also gives bad actors a larger area to attack.

This surface includes vital infrastructure, which is particularly susceptible to cyberattacks, according to DARPA specialists, because there aren’t enough technologies to adequately secure systems on a large scale. Cyber defenders face a formidable attack surface, which has been evident in recent years as malevolent cyber actors take advantage of this state of affairs to pose risks to society. Notwithstanding these weaknesses, contemporary technological advancements might offer a way around them.

AIxCC DARPA

According to Perri Adams, DARPA’s AIxCC program manager, “AIxCC represents a first-of-its-kind collaboration between top AI companies, led by DARPA to create AI-driven systems to help address one of society’s biggest issue: cybersecurity.” Over the past decade, exciting AI-enabled capabilities have emerged. DARPA see a lot of potential for this technology to be used to important cybersecurity challenges when handled appropriately. DARPA can have the most influence on cybersecurity in the nation and the world by automatically safeguarding essential software at scale.Image credit to DARPA

Participation at AIxCC will be available on two tracks: the Funded Track and the Open Track. Competitors in the Funded Track will be chosen from submissions made in response to a request for proposals for Small Business Innovation Research. Funding for up to seven small enterprises to take part will be provided. Through the competition website, Open Track contestants will register with DARPA; they will not receive funds from DARPA.

During the semifinal phase, teams on all tracks will compete in a qualifying event where the best scoring teams (up to 20) will be invited to the semifinal tournament. The top five scoring teams from these will advance to the competition’s final stage and win cash prizes. Additional cash awards will be given to the top three contestants in the final competition.

DARPA AIxCC

Leading AI businesses come together at DARPA AIxCC, where they will collaborate with DARPA to make their state-of-the-art technology and knowledge available to rivals. DARPA will work with Anthropic, Google, Microsoft, and Open AI to allow rivals to create cutting-edge a cybersecurity solutions.

As part of the Linux Foundation, the Open Source Security Foundation (OpenSSF) will act as a challenge advisor to help teams develop AI systems that can tackle important cybersecurity problems, like the safety of a crucial infrastructure and software supply chains. The majority of software, and therefore the majority of the code that requires security, is open-source and frequently created by volunteers in the community. Approximately 80% of contemporary software stacks, which include everything from phones and cars to electricity grids, manufacturing facilities, etc., use open-source software, according to the Linux Foundation.

Lastly, there will be AIxCC competitions at DEF CON and additional events at Black Hat USA. These two globally renowned cybersecurity conferences bring tens of thousands of practitioners, experts, and fans to Las Vegas every August. There will be two stages of AIxCC: the semifinal stage and the final stage. DEF CON in Las Vegas will hold 2024 and 2025 semifinals and finals.

DEF CON 32’s AI Village

Google will keep supporting the  AI Village at DEF CON, which informs attendees about security and privacy concerns related to AI, in addition to work with the AIxCC. In addition to giving the Village a $10,000 donation this year, Google will be giving workshop participants Pixel books so they can acquire practical experience with the nexus of cybersecurity and AI. After a successful red teaming event in which Google participated and security experts worked on AI security concerns, the AI Village is back this year.

Google excited to see what fresh ideas emerge from the AIxCC Semifinal Competition and to discuss how they may be used to safeguard the software that Google all use. Meanwhile, the Secure AI Framework and the recently established Coalition for Secure AI provide additional information about their efforts in the field of AI security.