In DARPA intend to use artificial intelligence to protect against existing and newest cyber threats
According to cybersecurity experts, hackers, on average, have 312 days a year to exploit zero-day vulnerabilities until they are identified and eliminated. In this regard, the management of the Pentagon Advanced Research Programs Office DARPA announced the launch of a new project, the purpose of which will be to determine the potential capabilities of artificial intelligence in detecting and eliminating vulnerabilities of a new software in a time calculated from a few seconds to several minutes.
As part of this new program provides a kind of competition Cyber ​​Grand Challenge, which is a real battle of artificial intelligence systems. As part of the task assigned to the AI ​​systems, it is planned to hack the server of the conditional adversary with simultaneous protection of its own servers, as well as to find and correct the flaws of the software “on the fly”. The winner of the competition - the author of the development of the system will receive a prize of $ 2 million, the second place holder claims for 1 million, the prize fund for the third place will be 750 thousand dollars.
The result of the battle of AI, according to the DARPA leadership, will allow to determine the artificial intelligence system, even if it needs some further refinement, but already today it can provide effective assistance to cybersecurity experts.
')
The “defense capability” of computing systems today is ensured thanks to the constant work of a group of highly qualified specialists, identifying signs of threats and developing measures to combat them in real time. Given the complexity and versatility of this process, as well as the number of error messages in case of software failures, each of which requires analysis, there is no need to speak of effective counteraction in the case of large-scale cyber attacks.
In this regard, the integration of AI capabilities will save experts from the enormous amount of routine work to monitor the current situation. Such systems, confident in the agency, will be able to perform faster and more comprehensive problem analysis, minimize reaction time, automatically generate corrective program codes that eliminate detected vulnerabilities.
The Cyber ​​Grand Challenge is scheduled for August 4, 2016 in Las Vegas. For 10 hours, members of development teams will only be able to watch how their creations battle with each other in a virtual digital space. But the key stage of the competition will be the final, in which artificial intelligence systems claiming the laurels of winners will operate within an isolated network, and a completely new and unknown software code will operate on the defendant servers. And at this stage in order to proceed to the detection and elimination of AI vulnerabilities, you will need to independently determine, study the network communication language and the logic of program interaction.
In order to attack competitors, the AI ​​system must independently detect the weak points of the adversary’s servers. When a vulnerability is detected, the system will need to inform the jury members of this and make an analytical forecast of the consequences that may arise if the detected vulnerability is used. Something like this acts as a billiard player who is planning a complex multi-strike blow with a few touches, which should result in a ball hitting a certain pocket. AI systems will not try to “knock down” competitors' servers, but will score points by detecting vulnerabilities in other servers, and lose points if they cannot protect their own servers, eliminating detectable vulnerabilities in time.
“After 21 days we are going to make a big scientific experiment in the field of artificial intelligence and at the same time an experiment to ensure the viability of computer systems. While we are not even ready to imagine what kind of results we will be able to get, ”wrote representatives of DARPA,“ But even a convincing victory in the Cyber ​​Grand Challenge competition will be only the beginning of a long path of development of artificial intelligence. The next step will be the performance of the winner at the conference of hackers DEF CON 2016, where AI will have to resist the actions of highly skilled and reputable expert hackers. ”
A source
That's all with you was Dronk.Ru. Do not forget to return money for purchases in China and subscribe to our blog , there will be many more interesting things.
We recommend:
- Save up to 8% on every purchase on AliExpress and other online stores in China
- Why do online stores give money for purchases?
- Return your money - Choose a cashback service for Aliexpress
- The history of the development of Dronk.ru - from choosing quadcopters to returning money for purchases on AliExpress and not only
- The best cashback service or 5 main criteria for evaluating cashback service
As part of this new program provides a kind of competition Cyber ​​Grand Challenge, which is a real battle of artificial intelligence systems. As part of the task assigned to the AI ​​systems, it is planned to hack the server of the conditional adversary with simultaneous protection of its own servers, as well as to find and correct the flaws of the software “on the fly”. The winner of the competition - the author of the development of the system will receive a prize of $ 2 million, the second place holder claims for 1 million, the prize fund for the third place will be 750 thousand dollars.
The result of the battle of AI, according to the DARPA leadership, will allow to determine the artificial intelligence system, even if it needs some further refinement, but already today it can provide effective assistance to cybersecurity experts.
')
The “defense capability” of computing systems today is ensured thanks to the constant work of a group of highly qualified specialists, identifying signs of threats and developing measures to combat them in real time. Given the complexity and versatility of this process, as well as the number of error messages in case of software failures, each of which requires analysis, there is no need to speak of effective counteraction in the case of large-scale cyber attacks.
In this regard, the integration of AI capabilities will save experts from the enormous amount of routine work to monitor the current situation. Such systems, confident in the agency, will be able to perform faster and more comprehensive problem analysis, minimize reaction time, automatically generate corrective program codes that eliminate detected vulnerabilities.
The Cyber ​​Grand Challenge is scheduled for August 4, 2016 in Las Vegas. For 10 hours, members of development teams will only be able to watch how their creations battle with each other in a virtual digital space. But the key stage of the competition will be the final, in which artificial intelligence systems claiming the laurels of winners will operate within an isolated network, and a completely new and unknown software code will operate on the defendant servers. And at this stage in order to proceed to the detection and elimination of AI vulnerabilities, you will need to independently determine, study the network communication language and the logic of program interaction.
In order to attack competitors, the AI ​​system must independently detect the weak points of the adversary’s servers. When a vulnerability is detected, the system will need to inform the jury members of this and make an analytical forecast of the consequences that may arise if the detected vulnerability is used. Something like this acts as a billiard player who is planning a complex multi-strike blow with a few touches, which should result in a ball hitting a certain pocket. AI systems will not try to “knock down” competitors' servers, but will score points by detecting vulnerabilities in other servers, and lose points if they cannot protect their own servers, eliminating detectable vulnerabilities in time.
“After 21 days we are going to make a big scientific experiment in the field of artificial intelligence and at the same time an experiment to ensure the viability of computer systems. While we are not even ready to imagine what kind of results we will be able to get, ”wrote representatives of DARPA,“ But even a convincing victory in the Cyber ​​Grand Challenge competition will be only the beginning of a long path of development of artificial intelligence. The next step will be the performance of the winner at the conference of hackers DEF CON 2016, where AI will have to resist the actions of highly skilled and reputable expert hackers. ”
A source
That's all with you was Dronk.Ru. Do not forget to return money for purchases in China and subscribe to our blog , there will be many more interesting things.
We recommend:
- Save up to 8% on every purchase on AliExpress and other online stores in China
- Why do online stores give money for purchases?
- Return your money - Choose a cashback service for Aliexpress
- The history of the development of Dronk.ru - from choosing quadcopters to returning money for purchases on AliExpress and not only
- The best cashback service or 5 main criteria for evaluating cashback service
Source: https://habr.com/ru/post/396191/
All Articles