Artificial Immune System For Software Rejuvenation

With software growing in scale and complexity, software aging is becoming the newtrend threatening the health of the software system. Software system, like people, gets old.Software aging is a phenomenon that the state and capability of the software systemdegrades with time, affected by the fluctuation of the running environment and hiddensoftware defects. The occasional, uncertainty and environmental dependency features ofsoftware aging make the countermeasures hard to be executed effectively.The countermeasures to software aging, called software rejuvenation, employphilosophies based on either model or measurement. Although model based systems havethe advantage of preventing known aging, they are incapable of detecting novel agingbecause they ignore the actual running status of the software. Measure based approaches,can detect partial aging modes according to specific software behaviors. But they are proneto miss the hybrid, dynamic aging modes, resulting in a higher false negative (FN) rate. Inaddition, both of the methods are dependent on complete aging samples which limit theapplicability of them because aging occurs occasionally.Artificial Immune System (AIS) is an emerging interdisciplinary subject which uses thehuman immune system (HIS) that adaptively protects the health of the human bodies asinspirations. Since aging is a typical health problem (medically known as sub-health‘), webelieve that AIS can deal with the unknown aging modes effectively because it is adaptive,dynamic, diverse and does not rely on the abnormal samples. Therefore, this paper proposedan artificial immune model for software rejuvenation and completed the following work:(1) Construction of the framework of artificial immune model for software rejuvenation:the framework is composed of artificial innate immune system which perceives theaging and artificial adaptive immune system which recognizes and evaluates the aging.Artificial Antigen Presenting Cell (APC) and lymphocytes are key components for theinnate and adaptive immune systems respectively. By collaborating with each other,they constitute a more complete, adaptive system for software rejuvenation.(2) Construction of the artificial innate immune system for software aging perception:learning from the mechanism that APCs perceive danger from the changes of theinternal environment of the human bodies, this paper proposed artificial APC model.The model adaptively extracts danger signals from the subtle changes (in the form of digital differential‘) and their correlations in system environment. Then artificial APCsrecognize and fuse these danger signals to identify the health condition of the systemand discover the program that might be aged. In addition, artificial APCs can extractproper danger signals which fit current situation through evolution. This gives theartificial APCs abilities to cope with the dynamic feature of the software aging.(3) Construction of the artificial lymphocyte model for software aging recognition andevaluation: In human bodies, lymphocytes recognize specific non-self‘antigens and their concentration reflects the extent of infection. Inspired by this mechanism, thispaper proposed artificial lymphocyte model to recognizing and evaluating softwareaging. Guided by artificial innate immune system, artificial lymphocytes recognizespecific aging behaviors exhibited by suspected program to determine it is aged or not.If the recognition is successful, the concentration of according artificial lymphocyteincreases and vice versa attenuates. By measuring the concentration, we can evaluatethe extent of the aging. The cooperation between this model and artificial innateimmune system not only reduces the false alarm rate when detecting novel aging, butalso identifies the specific program that is aged.(4) Example Validation of software rejuvenation: this paper focuses on the validation of thefeasibility of the method we proposed. Detection experiments were carried out with aprototype after choosing typical instances of software aging. Experimental results showthat the prototype can detect hidden aging in the system with low FP rate and FN rate.The digital differential can adapt to the changes of environment of the monitoredsystem and produce proper danger signals. Through evolution, artificial APCs canchoose proper combinations of danger signals that fit to the current situation, andreduce the FP rate.Using the human immune system (HIS) that adaptively protects the health of the humanbodies as inspirations, this paper proposed artificial immune model for softwarerejuvenation. The artificial APCs in the model perceive aging from the subtle changes andtheir correlations in system environment adaptively. Then the artificial lymphocytes in themodel are used to recognize and evaluate the aging. Compared to other softwarerejuvenation methods, this model can not only detect unknown aging adaptively, but alsoremember the aging patterns emerged before. It enhances the performance when detectingthe unknown, uncertain aging.