EvoAttack-testing operating systems by learning harmful system call sequences

@mastersthesis{citeulike:271385, abstract = {Operating systems testing is vital to the security of all information processing systems. System security has been traditionally an ongoing struggle between operating systems designers and the hacking community. While testing should be both methodical and complete, the sheer scope of the task makes both goals rather impossible to achieve. Human creativity within the hacking community pushes software into areas that the designers did not anticipate, thus revealing weaknesses. Evolutionary algorithms offer designers a unique way to examine the viability of their code. Evolutionary search techniques, because of the use of randomness as well as direction based on evaluation, help to simulate some aspects of the human creative process, mimicking hackers. In this work we show that already rather simple evolutionary searches allow us to find weaknesses in the test operating system, resulting in a crash of the system and the necessity to reboot\—a clear system flaw.}, author = {Williams, Timothy R. }, citeulike-article-id = {271385}, keywords = {se, security, testing}, posted-at = {2005-08-02 05:13:29}, priority = {2}, school = {UNIVERSITY OF CALGARY (CANADA)}, title = {EvoAttack-testing operating systems by learning harmful system call sequences}, url = {http://wwwlib.umi.com/dissertations/fullcit/MQ93411}, year = {2004} }

TY - THES ID - citeulike:271385 L3 - citeulike-article-id:271385 TI - EvoAttack-testing operating systems by learning harmful system call sequences N2 - Operating systems testing is vital to the security of all information processing systems. System security has been traditionally an ongoing struggle between operating systems designers and the hacking community. While testing should be both methodical and complete, the sheer scope of the task makes both goals rather impossible to achieve. Human creativity within the hacking community pushes software into areas that the designers did not anticipate, thus revealing weaknesses. Evolutionary algorithms offer designers a unique way to examine the viability of their code. Evolutionary search techniques, because of the use of randomness as well as direction based on evaluation, help to simulate some aspects of the human creative process, mimicking hackers. In this work we show that already rather simple evolutionary searches allow us to find weaknesses in the test operating system, resulting in a crash of the system and the necessity to reboot—a clear system flaw. KW - se KW - security KW - testing AU - Williams, Timothy PY - 2004/// UR - http://wwwlib.umi.com/dissertations/fullcit/MQ93411 ER -