International Journal of Security and Its Applications Vol. 10, No. 4 (2016), pp.249-266 http://dx.doi.org/10.14257/ijsia.2016.10.4.24
An Effective Approach for Classification of Advanced Malware with High Accuracy * Ashu Sharma1 and Sanjay K. Sahay2, 1
Research scholar, Department of Computer Science and Information System, Assistant Professor, Department of Computer Science and Information System, 1,2 Birla Institute of Technology and Science, K. K. Birla Goa Campus, NH-17B, By Pass Road, Zuarinagar-403726, Goa, India {1p2012011, 2ssahay}@goa.bits-pilani.ac.in 2
Abstract Combating malware is very important for software/systems security, but to prevent the software/systems from the advanced malware, viz. metamorphic malware is a challenging task, as it changes the structure/code after each infection. Therefore in this paper, we present a novel approach to detect the advanced malware with high accuracy by analyzing the occurrence of opcodes (features) by grouping the executables. These groups are made on the basis of our earlier studies [1] that the difference between the sizes of any two malware generated by popular advanced malware kits viz. PS-MPC, G2 and NGVCK are within 5 KB. On the basis of obtained promising features, we studied the performance of thirteen classifiers using N-fold cross-validation available in machine learning tool WEKA. Among these thirteen classifiers we studied in-depth top five classifiers (Random forest, LMT, NBT, J48 and FT) and obtain more than 96.28% accuracy for the detection of unknown malware, which is better than the maximum detection accuracy (~95.9%) reported by Santos et al (2013). In these top five classifiers, our approach obtained a detection accuracy of ∼97.95% by the Random forest. Keywords: Anti-Malware, Static Analysis, WEKA, Machine Learning, Decision Tree
1. Introduction “Malware refers to a program that is inserted into a system, usually covertly, with the intent of compromising the confidentiality, integrity, or availability of the victim’s data, applications, or operating system or of otherwise annoying or disrupting the victim” [2]. From the last four decades, malware is continuously evolving with high complexity to evade the available detection technique. These malware are basically classified as first and second generation malware. In the first generation, structure of the malware does not change, while in the second generation, structure changes to generate a new variant, keeping the action same [3]. On the basis of how variances are created in malware, second generation malware are further classified into Encrypted, Oligormorphic, Polymorphic and Metamorphic Malware [4]. These malware changes its structure in random and unpredictable ways each time it replicates, hence hard to detect. According to McAfee technical report of 2014, there are more than 200 million known malware samples [5]. The Symantec 2014 Internet Security Threat report states that 2013 was the mega breach year [6] (~62% more breaches then 2012). The F-secure document reported an increase in malware attacks against mobile devices based on Android and Apple iOS [7]. This increase in threat from malware is due to wide spread use of World Wide Web. An estimate shows that the web-based attacks were increased 36% with over 4,500 new *
[email protected]
ISSN: 1738-9976 IJSIA Copyright ⓒ 2016 SERSC
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
attacks each day, annoying/disrupting the victim in terms of confidentiality, integrity, availability of the victims data etc. [8]. The malware attacks/threat are not only limited to individual level, but there are state sponsored highly skilled hackers developing customized malware to disrupt industries and for military espionage [9]. Such attacks can alter the operation of industrial systems, disrupt power plants, e.g. the StuxNet and Duqu malware [10]. The intrusion into Google’s systems demonstrates how well-organized attacks are designed to maintain long-term access to an organization’s network [11]. To combat threats/attacks from the malware, signature-based software (anti-malware) were widely deployed. However, its an indisputable fact that these traditional approach of combating the threats/attack with a technology-centric are ineffective to detect today’s highly sophisticated customized malware. Hence attacks from such malware to the computing world are increasing every day. The consequence will be devastating if in time adequate measures had not been taken. Therefore, there is a need that both academia and anti-malware developers should continually work to combat the threats/attacks from the evolving malware. The most popular techniques used for the detection of malware are signature matching, heuristics-based detection, malware normalization, machine learning, etc. [4]. In recent years, machine learning techniques are studied by many authors and proposed different approaches [12] [13] [14], which can supplement traditional antimalware system. For the detection of malware by machine learning technique, feature selection plays a vital role. In the literature, many feature selection approaches are discussed viz. Olivier Henchiri et al. 2006 [15], Siddiqui et al. 2008 [16], B. Mehdi et al. 2009 [17] and Santos et al. 2013 [18] used hierarchical, unary variable removal method, Goodness evaluator and Weighted Term Frequency (WTF) respectively for the feature selection. The maximum accuracy they obtained was 95.26%. In this paper, our approach outperforms the accuracy obtained by these authors by more than ∼2%. The paper is organized as follow, in next section related work is discussed and in section 3 we present our approach, The section 4 discuss the experimental results and finally section 5 contains the conclusion of the paper.
2. Related Work The first virus was created in 1970 [19] and since then there is a strong contest between the attackers and defenders. This rat-race led to the improvement in both malware and its detection techniques. To defend the malware attacks, anti -malware groups are developing new techniques. On the other hand, malware developers are adopting new tactics/methods to avoid the malware detectors. Initially, the tools and techniques of malware analysis were in the domain of anti-malware vendors. However, the use of malware for espionage, sophisticated cyber-attacks and other crimes motivated the academicians and digital investigators to develop an advanced method to combat the threats/attacks from it. In the year 2001, Schultz et al. [20] were the first to introduce the concept of data mining for detecting malware. They used three different static features for malware classification (Portable Executable, strings and byte sequences). Kolter and Maloof (2004) evaluated data sets using Instance-Based Learning Algorithms (IBK), TF-IDF, Naive Bayes, Support Vector Machine (SVM) and Decision tree [21]. Among the classifiers used by them, Decision tree outperformed. In the year 2005, Karim et al. [22] addressed the tracking of malware evolution based on opcode sequences and permutations. O. Henchiri et al. (2006) proposed a hierarchical feature extraction algorithm and used ID3, j48, Naive Bayes and SMO classifier and obtained the maximum of 92.56% accuracy [15]. In the year 2007, Bilar uses small dataset (67 malware and 20 benign samples) to examine the difference in opcode frequency distribution between malicious and benign programs [23]. He found that malware opcode distribut ion differs significantly from benign programs and also observed that some opcodes
250
Copyright ⓒ 2016 SERSC
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
seen to be a stronger predictor of frequency variation. He however, did not apply it for the classification of advanced malware. In the year 2008, Tian et al. particularly classified Trojan malware using function length frequency [24]. Their results indicate that the function length along with its frequency is significant in identifying malware family and can be combined with other features for fast and scalable malware classification. In the year 2008, Siddiqui et al. [16] used variable length instruction sequence for detecting worms in the wild. They tested their method on a data set of 2774 (1444 worms and 1330 benign files) and got 95.6% detection accuracy. In the year 2008, Moskovitch et al. [25] [26] compared the different classifiers by byte-sequence n-grams (3, 4, 5 or 6). Among the classifiers they studied BDT, DT and ANN outperformed NB, BNB and SVM classifiers, exhibiting lower false positive rates. S. Momina Tabish (2009) used AdaBoostM1 algorithm for classification by taking n-gram frequency as a feature and reported 90% detection accuracy [27]. In the year 2010, Bilal Mehdi et. al. [28] used hyper-grams (generalized n-gram) and obtained 87.85% detection accuracy and claimed no false alarm. Santos et al. in the year 2011 pointed out that supervised learning requires a significant amount of labeled executables for both malware and benign programs, which is difficult to obtain, hence they proposed a semi-supervised learning method for detecting unknown malware, which does not require a large amount of labeled data [29]. They obtained 86% of accuracy by labeling only 50% of the selected data set. In the subsequent paper [18] in 2013, they used Term Frequency for modeling different classifiers and found that SVM outperforms with accuracy of 92.92% and 95.90% respectively for one opcode and two opcode sequence length respectively. Recently in 2014, Kevin Allix et al. [12] took a size-able dataset of over 50,000 android applications and implemented using 4 well-known machine learning algorithms (RandomForest, J48, JRip and LibSVM) with ten-fold cross-validation. He claimed his approach outperformed existing machine learning approaches, however on usual size datasets performance does not translate in the wild.
3. Our Approach In order to uncover the unknown malware with high accuracy, our novel approach as shown in Figure 1 involves finding the promising features (Algo. 1), training of classifiers and detection of unknown malware.
3.1. Building the Datasets and Feature Selection To build the datasets, we downloaded 11088 malware from malacia-project [30] and collected 4006 benign programs (cross checked from virustotal.com [31]) from different systems.
Copyright ⓒ 2016 SERSC
251
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
Malwares collected from malicia project and benign data collected from uninfected computer systems
Malware
Benign
programs
programs
Preprocessing of data
Training data
Testing data
Building the datasets
Feature Selection algorithm
Machine Learning
Top 5 classifiers
technique on
based on accuracy
13 classifiers
(Random forest, LMT, NBT, J48, FT)
Set of features
Results
Figure 1. Flow Chart for the Detection of Unknown Malware
Algorithm 1: Feature Selection INPUT: Pre-processed data of all groups, Nb→ Number of benign executables and Nm→ Number of malware executables OUTPUT: List of features. BEGIN for all groups GK do for all ith benign executable in the group GK do Compute the normalized frequency FKb of each opcode oj
F K (o ) = (å f (o )) / N j
b
i
j
b
end for for all ith malware executable in the group GK do Compute the normalized frequency FKm of each opcode oj
F K (o ) = (å f (o )) / N m
j
i
j
m
end for for all opcode oj do Subtract the frequencies Fkb and Fkm.
D K (o ) = F K (o ) - F K (o ) j
b
j
m
j
end for Sort the obtained DK(o). end for Set a threshold on Dk(o) to select the promising opcodes features such that from each Gk at least 10 opcodes get selected.
252
Copyright ⓒ 2016 SERSC
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
return Union of the selected features of all the groups.
The promising features of the executables are obtained by clubbing the dataset in 5 KB size of 100 groups [1] as in the collected dataset ~97.18% malware are below 500 KB (Figure 2) and the difference between the sizes of any two malware generated by popular advanced malware kits viz. NGVCK [32], PS-MPC [33] and G2 [34] are within 5 KB. Hence, the features obtained will have a signature of maximum executables to detect the unknown malware. Our features are opcodes of the executables obtained by objdump utility available in the Linux system. To identify the each opcode we labeled it with a fixed integer as given in the Appendix A. To differentiate malware and benign programs we obtained the features as given in algo. 1. 500 450
Number of Malware ®
400 350 300 250 200 150 100 50 0 0
200
400
600
800
1000
1200
1400
1600
Size in Kilobytes ®
Figure 2. Number of Malware with Respect to File Size
3.2. Training of the Classifiers To find the best classifiers for detection of unknown malware, we investigated thirteen tree based classifiers viz. Random forest, J48, REPTREE, LMT, Decision stump, ADT, NBT, FT, LAD, Random Tree, Simple CART, BFT and J48 Graft available in the popular and widely used suite of machine learning software known as WEKA (a collection of visualization tools and algorithms for data analysis and predictive modeling, together with graphical user interfaces for easy access to this functionality). Then with the obtained features, we run the WEKA n-fold cross-validation to train all the selected classifiers. Figure 3 shows the accuracy obtained by all classifiers for n = 2,4,6...,16 folds. We observed that Random forest is the best classifier and its accuracy is almost flat after n = 2. Rest twelve classifiers accuracy fluctuates, however after ten-fold cross-validation the fluctuations are least and we observe maximum correctness in the accuracy.
Copyright ⓒ 2016 SERSC
253
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
1 0.99 0.98 0.97
Accuracy ®
0.96 0.95 0.94
randomforest j48 reptree lmt decisionstump adt nbt ft lad randomtree simplecart bft j48graft
0.93 0.92 0.91 0.9 0.89 0.88 0.87 2
4
6
8
10
12
14
16
Folds ®
Figure 3. Accuracy of the Thirteen Classifiers with N-Fold Cross Validation
3.3 Detection of Unknown Malware Among the thirteen studied classifier, we selected top five (Random forest, LMT, NBT, J48 and FT) for in-depth analysis. To study the overall performance of these classifiers, we randomly selected 750 malware and 610 benign programs from all the groups, such that at least five executables from each group can be randomly tested by the trained classifiers with ten-fold cross-validation for the detection of unknown malware.
4. Experimental Results
1
Normalized Frequency ®
0.8
Benign Malware
0.6
0.4
0.2
2 6 7 12 50 59 74 75 78 79 80 101 104 179 418 426 431 432 441 448 451 454 457 461 473 479 482 494 508 514 515 525 530 531 538 555 567 578 624 625 632 639 692 737 743 781 834 839 887 888 889 891 893 906 1128 1135 1136
0
Opcode ®
Figure 4. Normalized Opcode Occurrence of all the Collected Malware and Benign Program Keeping Threshold 0.02
254
Copyright ⓒ 2016 SERSC
2 3 4 5 6 7 12 18 19 50 59 61 74 75 78 79 80 81 84 101 104 114 115 160 161 179 181 193 194 204 324 325 377 386 387 418 420 426 431 432 434 437 438 440 441 442 445 448 451 454 457 461 464 467 470 473 476 479 482 485 488 491 494 497 503 506 508 514 515 518 525 530 531 538 539 542 549 555 567 578 588 593 624 625 632 639 640 641 692 693 695 737 738 740 743 764 765 773 774 781 828 834 839 877 878 887 888 889 891 893 906 1128 1135 1136
Normalized Frequency ®
7
2
Copyright ⓒ 2016 SERSC 0.8
1136
1128
906
893
834
782
781
743
737
693
692
641
640
0.8
639
632
625
624
607
603
599
589
578
571
567
514
494
482
480
479
476
470
467
461
454
451
441
438
437
432
431
426
387
386
325
324
179
104
101
75
74
50
12
Normalized Frequency ®
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
1
Benign Malware
0.6
0.4
0.2
0
Opcode ®
Figure 5. Normalized Opcode Occurrence of the Malware and Benign Program of Size 10-15 KB Keeping Threshold 0.02 1
Benign Malware
0.6
0.4
0.2
0
Opcode ®
Figure 6. Normalized Opcode Occurrence of the Malware and Benign Program of Size 140-145 KB Keeping Threshold 0.02
255
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
1
Benign Malware
Normalized Frequency ®
0.8
0.6
0.4
0.2
2 3 6 7 12 50 59 74 75 78 79 80 101 104 126 127 159 160 161 179 193 194 204 207 208 426 431 432 434 437 438 441 451 454 461 467 479 482 494 503 508 514 525 530 531 538 549 555 567 571 578 604 624 625 632 639 640 641 692 693 737 738 740 743 781 834 839 877 878 888 889 891 893 906 1128 1136
0
Opcode ®
Figure 7. Normalized Opcode Occurrence of the Malware and Benign Program of size 240-245 KB Keeping Threshold 0.02 1
Benign Malware
Normalized Frequency ®
0.8
0.6
0.4
0.2
1136
906
1128
893
864
834
781
743
737
692
639
632
578
567
514
482
479
461
454
441
432
431
426
179
104
75
101
74
50
7
12
2
0
Opcode ®
Figure 8. Normalized Opcode Occurrence of the Malware and Benign Program of size 415-420 KB Keeping Threshold 0.02
256
Copyright ⓒ 2016 SERSC
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
Figure 9. Difference in the Occurrence of Respective Opcodes between Benign and Malware Program of Size 0-5 KB and 15-20 KB Keeping Threshold 0.02
Figure 10. Difference in the Occurrence of Respective Opcodes between Benign and Malware Program of Size 30-35 KB and 55-60 KB Keeping Threshold 0.02 As discussed in section 3, we disassembled all the malware and benign programs of below 500 KB size. Then we computed the normalized opcode occurrence of all malware and benign programs, for each group separately. We observed that the opcode occurrence in the malware and benign programs differ in large. Fig. 4 shows the normalized opcode occurrence of all the malware and benign programs and Fiureg 5, 6, 7, & 8 shows the
Copyright ⓒ 2016 SERSC
257
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
normalized opcode occurrence for the group 10-15, 140-145, 240-245 & 415-420 KB size respectively keeping the lower threshold 0.02. To find the dominant features of malware and benign programs we computed the difference in opcode occurrence between the benign and malware programs of each group and found that dominant opcodes vary from group to group. Figure 9. shows the difference in the occurrence of respective opcodes between benign and malware program of size 0-5 KB and 15-20 KB size keeping the lower threshold 0.02. Similarly, Figure 10 is plotted for the size 30-35 and 55-60 KB size of the data set. In the Figure 9, 10 upper side shows the opcodes that dominate in benign and lower side shows the opcode that dominates in malware. The effectiveness of the top five classifiers viz. Random forest Tree, LMT, NBT, J48 and FT has been studied with the randomly selected 750 malware and 610 benign programs. The analysis are done in WEKA with ten-fold cross-validation, in terms of True Positive Ratio (TPR), True Negative Ratio (TNR), False Positive Ratio (FPR), False Negative Ratio (FNR) and accuracy, defined as
TP TPR= TM ;
TN TNR= TB ;
FP FPR= TB ;
FN FNR= TM
TP+TN Accuracy= TM+TB×100 where, TP⟶ True positive, the number of malware correctly classified TN⟶ True negative, the number of benign correctly classified. FP⟶ False positive, the number of benign detected as malware. FN⟶ False negative, the number of malware detected as benign. TM⟶ Total number of malware. TB⟶ Total number of benign. From the analysis, it is clear that Random forest is the best classifier for identification of unknown malware. Nevertheless, the other classifiers are also reasonably good (> 96.2%) for the detection of unknown malware. The detail results obtained are shown in table 1. Table 1. Performance of the Top 5 Classifiers Classifiers True positive False negative Random forest 739 (98.53%) 11 (1.47%) LMT 734 (97.87%) 16 (2.13%) NBT 728 (97.07%) 22(2.93%) J48 729 (97.2%) 21 (2.8%) FT 729 (97.2%) 21 (2.8%)
False positive 6 (2.81%) 23 (4.04%) 19 (3.33%) 23 (4.04%) 8 (4.91%)
True Negative 554 (97.19%) 547 (95.96%) 551 (96.67%) 547 (95.96%) 542 (95.09%)
Accuracy 97.95% 97.04% 96.89% 96.66% 96.28%
We found that the classifier NBT, J48 and FT have the almost same True positive ratio. In these, the overall accuracy of Functional Tree classifier is lowest, which is basically due to high False positive ratio. Figure 11 shows the variation of False positive and False negative of the studied classifiers. We found that the False positives ratio of Random forest and LMT are almost double than False positive ratio, however for overall accuracy both has to be low. From Figure 12 we observed that the True positives of all classifiers are more biased towards the detection of malware. The obtained accuracy for all five classifiers are shown in Figure 13 and the comparison of our results with Santos et al., Siddiqui et al., Asaf Shabtai et. al. for Random forest and Mehdi et al., Santos et al., Olivier Henchiri et al. for J48 are shown in Figure 14. Among these authors our approach uncover the malware with the best accuracy (~97.95%).
258
Copyright ⓒ 2016 SERSC
International Journal of Security and Its Applications Vol. 10, No. 4 (2016)
5
Measures ®
4
3
2
FN FP 1 RandomForest
LMT
NBT
J48
FT
Classifiers ®
Figure 11. FP and FN of Top Five Classifiers 99
TP TN
Measures ®
98
97
96
95 RandomForest
LMT
NBT
J48
FT
Classifiers ®
Figure 12. TP and TN of Top Five Classifiers
Copyright ⓒ 2016 SERSC
259
International Journal of Security and Its Applications Vol. x, No. x, (20xx)
98
97.5
Accuracy ®
97
96.5
96
95.5
95 RandomForest
LMT
NBT
j48
FT
Classifiers ®
Figure 13. Accuracy of the Top Five Classifiers
Figure 14. Comparison of the Accuracy Obtained by our Approach and Others
5. Conclusion Traditional approach i.e. updating the signature database to combat advanced malware is ineffective. Therefore, in this paper, we presented a novel approach to detect the advanced malware with high accuracy. For the classification, we obtained the promising features (opcodes) by grouping the executables in 5 KB size. Extensive experiment has been done to study the performance of the classifiers viz. Random forest, LMT, NBT, J48 and FT in terms of TPR, TNR, FPR, FNR and accuracy by analyzing 11688 malware
260
International Journal of Security and Its Applications Vol. x, No. x, (20xx)
downloaded from malicia-project and 4006 benign programs collected from different systems. By our approach all five classifiers are able to uncover unknown malware with greater than 96.28% accuracy, which is better than the detection accuracy (~95.9%) reported by Santos et. al. (2013). Among these classifiers, we found that Random forest is the best (∼97.95%) classifier to detect the unknown malware. Thus, our approach outperforms to detect the unknown malware and hence can be an effective technique to complement the signature based mechanism or dynamic analysis. In future, we will collect more malware and benign and will perform in-depth size analysis for the classification of unknown malware.
Appendix ID 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46
Opcode .byte (bad) aaa aad aam aas adc adcb adcl adcq adcw add addb addl addpd addps addq addr16 adcx addr32 addsubps addr32csrex.B addr32csrex.RXB addr32fsrex.RX addr32gsrex.R addr32gsrex.RXB addr32gsrex.W addr32gsrex.WRB addr32gsrex.WRX addr32rex addr32rex.B addr32rex.RX addr32rex.RXB addr32rex.WR addr32rex.WRB addr32rex.WRX addr32rex.WXB addr32rex.X addr64 addsd addss addsubpd addw aesdec aesdeclast aesenc
ID 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276
Opcode fcmovnbe fcmovne fcmovnu fcmovu fcom fcomi fcomip fcoml fcomp fcompl fcompp fcomps fcoms fcos fdecstp fdiv fdivl fdivp fdivr fdivrl fdivrp fdivrs fdivs femms ffree ffreep fiadd fiaddl ficom ficoml ficomp ficompl fidiv fidivl fidivr fidivrl fild fildl fildll fimul fimull fincstp finit fist fistl fistp
ID 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
Opcode je je,pn je,pt jecxz jecxz,pn jecxz,pt jg jg,pn jg,pt jge jge,pn jge,pt jl jl,pn jl,pt jle jle,pn jle,pt jmp jmpq jmpw jne jne,pn jne,pt jno jno,pn jno,pt jnp jnp,pn jnp,pt jns jns,pn jns,pt jo jo,pn jo,pt jp jp,pn jp,pt jrcxz jrcxz,pn jrcxz,pt js js,pn js,pt lahf
ID 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736
Opcode pmuludq pop popa popaw popf popfq popfw popl popq popw por prefetch prefetchnta prefetcht0 prefetcht1 prefetcht2 prefetchw psadbw pshufb pshufd pshuflw pshufw pslld psllq psllw psrad psraw psrld psrldq psrlq psrlw psubb psubd psubq psubsb psubsw psubusb psubusw psubw punpckhbw punpckhdq punpckhqdq punpckhwd punpcklbw punpckldq punpcklwd
ID 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966
261
Opcode vaddsd vaddsubpd vaddsubps vaesdec vandnpd vandnps vandpd vandps vcmpltpd vcmpngess vcmppd vcmpps vcmpsd vcmpss vcomisd vcomiss vcvtdq2pd vcvtdq2ps vcvtpd2dq vcvtpd2dqx vcvtpd2dqy vcvtpd2psx vcvtps2dq vcvtps2pd vcvtps2ph vcvtsd2si vcvtsd2ss vcvtsi2sd vcvtsi2ssl vcvtss2sd vcvtss2si vcvttpd2dq vcvttpd2dqy vcvttps2dq vcvttsd2si vcvttss2si vdivpd vdivps vdivsd vdivss vdpps verr verw vfmadd213pd vfmadd231pd vfmadd231ss
International Journal of Security and Its Applications Vol. x, No. x, (20xx)
47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107
aesenclast aesimc aeskeygenassist and andb andl andnpd andnps andpd andps andq andw arpl blcfill bound bsf bsr bswap bt btc btcl btl btr btrl bts btsl btsq call callq callw cbtw clc cld cli cltd cltq clts cmc cmova cmovae cmovb cmovbe cmove cmovg cmovge cmovl cmovle cmovne cmovno cmovnp cmovns cmovo cmovp cmovs cmp cmpb cmpeqsd cmpl cmpltpd cmpltsd cmpltss
262
277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
fistpl fistpll fisttp fisttpl fisttpll fisub fisubl fisubr fisubrl fld fld1 fldcw fldenv fldenvs fldl fldl2e fldl2t fldlg2 fldln2 fldpi flds fldt fldz fmul fmull fmulp fmuls fnclex fndisi(8087 fneni(8087 fninit fnop fnsave fnsaves fnsetpm(287 fnstcw fnstenv fnstenvs fnstsw fpatan fprem fprem1 fptan frndint frstor frstors frstpm(287 fs fstp1 fsave fscale fsfsrex fsfsrex.RX fsfsrex.WRB fsfsrex.WXB fsgsrex.W fsgsrex.WB fsgsrex.WR fsin fsincos fsqrt
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
lar lcall lcallq lcallw lddqu ldmxcsr lds lea leave leaveq leavew les lfs lgdt lgdtl lgs lidt lidtl ljmp ljmpq ljmpw lldt lmsw loopew,pt lock lockrex lockrex.B lockrex.WB lockrex.WR lockrex.X lockrex.XB lods loop loop,pn loop,pt loope loope,pn loope,pt loopel loopew loopew,pn loopl loopne loopne,pn loopne,pt loopnel loopnew loopw lret lretq lretw lsl lss ltr maskmovq maxps minpd minps minss montmul mov
737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797
push pusha pushaw pushf pushfq pushfw pushl pushq pushw pxor rcl rclb rcll rclq rclw rcpps rcpss rcr rcrb rcrl rcrq rcrw rdmsr rdpmc rdtsc rep repe repne repnz repnzrex.R repnzrex.RX repnzrex.RXB repnzrex.W repnzrex.WRX repnzrex.WRXB repnzrex.XB repz rueu repzcsrex.XB repzrex.WRB repzrex.WX repzrex.WXB repzrex.X repzrex.XB ret retq retw rex retnw rg rgb rex.R rex.R rex.RB rex.RB rex.RX rglpsz rex.RXB rguid rex.W riid
967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
vfnmsubpd vfrczsd vfrczss vgatherqpd vhaddpd vhaddps vhsubpd vhsubps vlddqu vmaskmovdqu vmaxpd vmaxps vmaxsd vmaxss vminpd vminps vminsd vminss vmload vmmcall vmovapd vmovaps vmovd vmovddup vmovdqa vmovdqu vmovhpd vmovhps vmovlpd vmovlps vmovmskpd vmovmskps vmovntdq vmovntpd vmovntps vmovq vmovsd vmovshdup vmovsldup vmovss vmovupd vmovups vmptrld vmptrst vmread vmulpd vmulps vmulsd vmulss vmwrite vorpd vorps vpackssdw vpacksswb vpackuswb vpaddb vpaddd vpaddq vpaddsb vpaddusb vpaddusw
International Journal of Security and Its Applications Vol. x, No. x, (20xx)
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168
cmpneqpd cmpneqps cmpnlepd cmppd cmpps cmpq cmpsb cmpsl cmpsq cmpsw cmpunordps cmpw cmpxchg cmpxchg8b comisd comiss cpuid cqto cs cvtpd2dq csrex.B csrex.R csrex.RB csrex.RXB csrex.W csrex.WB csrex.WR csrex.WRB csrex.WRX csrex.WX csrex.WXB csrex.X csrex.XB cvtdq2pd cvtdq2ps cvtpd2ps cvtpi2ps cvtps2pd cvtps2pi cvtsd2si cvtsd2ss cvtsi2sd cvtsi2sdq cvtsi2ss cvtss2sd cvttpd2dq cvttpd2pi cvttps2pi cvttsd2si cvttss2si cwtd cwtl daa das data16 data16addr32rex.WRB data16data16rex.R data16data16rex.WXB data16gsrex.RXB data16gsrex.WRB data16rex
338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398
fsrex fsrex.B fsrex.R fsrex.RB fsrex.RX fsrex.RXB fsrex.W fsrex.WB fsrex.WR fsrex.WRB fsrex.WRX fsrex.WRXB fsrex.WX fsrex.WXB fsrex.XB fst fstcw fstenv fstl fstp fstpl fstps fstpt fsts fstsw fsub fsubl fsubp fsubr fsubrl fsubrp fsubrs fsubs ftst fucom fucomi fucomip fucomp fucompp fwait fxam fxch fxrstor fxsave fxtract fyl2x fyl2xp1 getsec gs gdiplus gsaddr32rex.R gsaddr32rex.WXB gscsrex.RXB gsdata16rex gsdsrex.R gsdsrex.RXB gsdsrex.WRB gsdsrex.WRX gsesrex.W gsfsrex.B gsfsrex.RB
568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628
movabs movapd movaps movb movd movdq2q movdqa movdqu movhlps movhps movl movlhps movlpd movlps movmskps movntdq movnti movntps movntq movq movsb movsbl movsbq movsbw movsd movsl movsldup movslq movsq movss movsw movswl movswq movupd movups movw movzbl movzbq movzbw movzwl movzwq movzww mul mulb mull mulpd mulps mulq mulsd mulss mulw neg negb negl negq negw nop nopq nopl nopw not
798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858
rex.WB rorx rex.WR roundpd rex.WRB roundps rex.WRX roundsd rex.WRXB rsldt rex.WX rsqrtss rex.WXB rsts rex.X rex.RXB rex.XB rueu rol rolb roll rolq rolw ror rorb rorl rorq rorw rsm rsqrtps sahf sar sarb sarl sarq sarw sbb sbbb sbbl sbbq sbbw scas seta setae setb setbe sete setg setge setl setle setne setno setnp setns seto setp sets sgdt sgdtl shl
1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088
263
vpaddw vpand vpandn vpavgb vpavgw vpcmpeqb vpcmpeqd vpcmpeqw vpcmpgtb vpcmpgtd vpcmpgtw vpcomud vpcomw vpextrw vphaddbd vphaddbw vphaddubq vphaddubw vphadduwq vphaddwq vphsubbw vphsubdq vphsubwd vpinsrw vpmacsdqh vpmacsdql vpmacssdql vpmacsswd vpmacssww vpmacswd vpmaddwd vpmaxsw vpmaxub vpminsw vpminub vpmovmskb vpmulhuw vpmulhw vpmulld vpmullw vpmuludq vpor vpperm vprotb vprotd vpsadbw vpshab vpshld vpshlw vpshufb vpshufhw vpshuflw vpslld vpsllq vpsllw vpsrad vpsraw vpsrld vpsrlq vpsrlw vpsubb
International Journal of Security and Its Applications Vol. x, No. x, (20xx)
169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229
data16rex.B data16rex.R data16rex.W data16rex.WB data16rex.WRB data16rex.WRX data16rex.WRXB data16rex.WXB data16rex.XB data32 dec decb decl decq decw div divb divl divpd divps divq divsd divss divw d3d8 ds dsrex dsrex.R dsrex.RXB dsrex.WB dsrex.WR dsrex.WRB dsrex.WRX dsrex.WRXB emms enter enterq enterw es encls esrex.RB esrex.RX esrex.W esrex.WB esrex.WRX esrex.X esrex.XB extrq f2xm1 fabs fadd faddl faddp fadds fbld fbstp fchs fclex fcmovb fcmovbe fcmove
264
399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
gsfsrex.W gsfsrex.WR gsfsrex.WRB gsgsrex.R gsgsrex.W gsrex gsrex.B gsrex.R gsrex.RB gsrex.RXB gsrex.W gsrex.WB gsrex.WR gsrex.WRB gsrex.WRX gsrex.WRXB gsrex.WXB gsrex.X gsrex.XB hlt hsubps icebp idiv idivb idivl idivq idivw imul imulb imull imulq imulw in inc incb incl incq incw insb insl insw int int3 into invd invlpg iret iretq iretw ja ja,pn ja,pt jae jae,pn jae,pt jb jb,pn jb,pt jbe jbe,pn jbe,pt
629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689
notb notl notw or orb orl orpd orps orq orw out outsb outsl outsw packssdw packsswb packuswb paddb paddd paddq paddsb paddsw paddusb paddusw paddw palignr pand pandn pause pavgb pavgw pcmpeqb pcmpeqd pcmpeqw pcmpgtb pcmpgtd pcmpgtw pcmpistri pextrw pf2id pf2iw pfcmpeq pfcmpgt pfrsqit1 phaddbd phaddd phadduwq phsubbq pi2fd pi2fw pinsrb pinsrw pmaddubsw pmaddwd pmaxsw pmaxub pminsw pminub pmovmskb pmulhuw pmulhw
859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919
shlb shld shll shlq shlw shr shrb shrd shrl shrq shrw shufps sidt sidtl sldt smsw sqrtpd sqrtps ss sz ssfsrex.W ssrex.B ssrex.RXB ssrex.WB ssrex.WRB ssrex.WX ssrex.WXB ssrex.X stc std sti stmxcsr stos str sub subb subl subpd subps subq subsd subss subw syscall sysenter sysexit sysret test testb testl testq testw ucomisd ucomiss ud1 ud2 unpckhpd unpckhps unpcklpd unpcklps vaddpd
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
vpsubq vpsubsb vpsubsw vpsubusb vpsubusw vpsubw vpunpckhbw vpunpckhdq vpunpckhqdq vpunpckhwd vpunpcklbw vpunpckldq vpunpcklqdq vpunpcklwd vrcpps vrcpss vrsqrtps vrsqrtss vshufpd vshufps vsqrtpd vsqrtps vsqrtsd vsqrtss vsubpd vsubps vsubsd vsubss vucomisd vucomiss vunpckhps vunpcklpd vunpcklps vxorpd vxorps vzeroupper wbinvd wrmsr xadd xchg xcrypt-cbc xcrypt-cfb xcrypt-ctr xcrypt-ecb xcrypt-ofb xgetbv xlat xor xorb xorl xorpd xorps xorq xorw xrstor xsave xsaveopt xsha1 xsha256
International Journal of Security and Its Applications Vol. x, No. x, (20xx)
230
fcmovnb
460
jcxz
690
pmullw
920
vaddps
Acknowledgments Mr. Ashu Sharma is thankful to BITS, Pilani, K.K. Birla Goa Campus for the support to carry out his work through Ph.D. scholarship No. Ph603226/Jul. 2012/01. We are also thankful to IUCAA, Pune for providing hospitality and computation facility where part of the work was carried out.
References [1]
[2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]
[13] [14]
[15]
[16] [17]
[18]
[19] [20] [21]
[22] [23] [24] [25]
A. Sharma, S. K. Sanjay, “Improving the detection accuracy of unknown malware by partitioning the executables in groups”, Proceedings of the 9th international confrence on advanced computing and comunication technologies, Nagpur, India, (2015). P. Mell, K. Kent, J. Nusbaum, “Guide to malware incident prevention and handling”, US Department of Commerce, Technology Administration, National Institute of Standards and Technology, (2005). Govindaraju, “Exhaustive statistical analysis for detection of metamorphic malware”, Master’s thesis, San Jose State University (2010). A. Sharma, S. K. Sahay, “Evolution and detection of polymorphic and metamorphic malware: A survey”, International Journal of Computer Applications., vol. 90, no. 2, (2014), pp 7-11. V. Weafer, Editor, “Mcafee labs threats report june 2014”, McAfee Labs threat report, (2014). P. Wood, Editor, “Internet security threat report 2014”, Symantec Corporation, USA, (2014). M. Hypp nen, “Threat report H1 2013”, F-Secure labs, Malaysia, (2013). P. Wood, Editor, “Internet security threat report”, Symantec Corporation, USA (2012). R. Stone, “A call to cyber arms”, Science., vol. 339, no. 6123, (2013), pp. 1026–1027. B. Bencsáth, G. Pék, L. Buttyán, M. Félegyházi, “Duqu: A stuxnet-like malware found in the wild”, CrySyS Lab, Hungary, (2011). M. K. Daly, “Advanced persistent threat”, Proceedings of the 23rd Large Installation System Administration Conference, Usenix, Baltimore, USA, vol. 35, no. 1 (2009), pp. 104-105. K. Allix, T. F. Bissyandé, Q. Jérome, J. Klein, Y. Le Traon, et al., “Large-scale machine learning-based malware detection: confronting the 10-fold cross validation scheme with reality”, Proceedings of the 4th ACM conference on Data and application security and privacy, ACM, (2014), pp. 163–166. E. Gandotra, D. Bansal, S. Sofat, “Malware analysis and classification: A survey”, Journal of Information Security., vol. 5, (2014), pp. 56-64. H. V. Nath, B. M. Mehtre, “Static malware analysis using machine learning methods”, Proceedings of the Recent Trends in Computer Networks and Distributed Systems Security, Springer Berlin Heidelberg, (2014), pp. 440–450. O. Henchiri, N. Japkowicz, “A feature selection and evaluation scheme for computer virus detection”, Proceedings of the Sixth International Conference on Data Mining, IEEE, Hong Kong, (2006), pp. 891– 895. M. Siddiqui, M. C. Wang, J. Lee, “Detecting internet worms using data mining techniques”, Journal of Systemics, Cybernetics and Informatics., vol. 6, no. 6, (2009), pp. 48–53. S. B. Mehdi, A. K. Tanwani, M. Farooq, “Imad: in-execution malware analysis and detection”, Proceedings of the 11th Annual conference on Genetic and evolutionary computation, ACM, Montreal, (2009), pp. 1553–1560. I. Santos, F. Brezo, X. Ugarte-Pedrero, P. G. Bringas, “Opcode sequences as representation of executables for data-mining-based unknown malware detection”, Information Sciences., vol 231 (2013) pp. 64–82. K. Gettman, Editor, “The art of computer virus research and defense”, Pearson Education, (2005). M. G. Schultz, E. Eskin, E. Zadok, S. J. Stolfo, “Data mining methods for detection of new malicious executables”, Proceedings of the IEEE Symposium on Security and Privacy, IEEE, (2001), pp. 38–49. Z. Kolter, M. A. Maloof, “Learning to detect malicious executables in the wild”, Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining, ACM, (2004), pp. 470–478. M. E. Karim, A. Walenstein, A. Lakhotia, L. Parida, “Malware phylogeny generation using permutations of code”, Journal in Computer Virology., vol. 1, (2005), pp. 13–23. D. Bilar, “Opcodes as predictor for malware”, International Journal of Electronic Security and Digital Forensics., vol. 1, no. 2, (2007), pp. 156–168. R. Tian, L. M. Batten, S. Versteeg, “Function length as a tool for malware classification”, Proceding of the 3rd International Conference on Malicious and Unwanted Software, IEEE, (2008), pp. 69–76. R. Moskovitch, D. Stopel, C. Feher, N. Nissim, Y. Elovici, “Unknown malcode detection via text categorization and the imbalance problem”, Proceedings of the IEEE International Conference on Intelligence and Security Informatics, IEEE, (2008), pp. 156–161.
265
International Journal of Security and Its Applications Vol. x, No. x, (20xx)
[26] R. Moskovitch, C. Feher, N. Tzachar, E. Berger, M. Gitelman, S. Dolev, Y. Elovici, “Unknown malcode detection using opcode representation”, Proceedings of the Springer on Intelligence and Security Informatics, (2008), pp. 204–215. [27] S. M. Tabish, M. Z. Shafiq, M. Farooq, “Malware detection using statistical analysis of byte-level file content”, Proceedings of the ACM SIGKDD Workshop on CyberSecurity and Intelligence Informatics, ACM, (2009), pp. 23–31. [28] B. Mehdi, F. Ahmed, S. A. Khayyam, M. Farooq, “Towards a theory of generalizing system call representation for in-execution malware detection”, Proceedings of the IEEE International Conference on Communications (ICC), IEEE, (2010), pp. 1–5. [29] I. Santos, J. Nieves, P. G. Bringas, “Semi-supervised learning for unknown malware detection”, Proceedings of the International Symposium on Distributed Computing and Artificial Intelligence, Springer, (2011), pp. 415–422. [30] A. Nappa, M. Z. Rafique, J. Caballero, “Driving in the cloud: An analysis of drive-by download operations and abuse reporting”, Proceedings of the Detection of Intrusions and Malware, and Vulnerability Assessment, Springer Berlin Heidelberg, (2013), pp. 1–20. [31] J. Canto, M. Dacier, E. Kirda, C. Leita, “Large scale malware collection: lessons learned”, Proceedings of the 27th International Symposium on Reliable Distributed Systems and Experiment Measurements on Resilience of Distributed Computing Systems, (2008). [32] A. Venkatesan, “Code obfuscation and virus detection”, Ph.D. thesis, San Jose State University (2008). [33] P. Beaucamps, “Advanced metamorphic techniques in computer viruses”, Proceedings of the International Conference on Computer, Electrical, and Systems Science, and Engineering-CESSE’07, (2007). [34] T. H. Austin, E. Filiol, S. Josse, M. Stamp, “Exploring hidden markov models for virus analysis: a semantic approach”, Proceedings of the 46th Hawaii International Conference on System Sciences, IEEE, Havaii, (2013), pp. 5039–5048.
Authors Ashu Sharma is a full time research scholar in Department of Computer Science and Information Systems, BITS-Pilani, K.K. Birla Goa Campus, India and perusing for his Ph.D degree on malware analysis under the supervision of Dr. Sanjay K. Sahay. He has published couple of papers in malware analysis and two papers are accepted in conference proceedings and will appear in Springer Verlag and Elsevier Procedia CS. Sanjay K. Sahay received his Ph.D in 2003 and currently working as an Assistant Professor in Computer Science and Information System at BITS-Pilani, K.K. Birla Goa Campus, India. His research interest includes malware analysis, data mining, gravitational waves and machine learning. He has published papers on topics like malware analysis, gravitational waves, machine learning and data mining.
266
