Скачать презентацию Hybrid Intrusion Detection with Weighted Signature Generation over Скачать презентацию Hybrid Intrusion Detection with Weighted Signature Generation over

d72d387f3d8a010f6c7540725fe3f11f.ppt

  • Количество слайдов: 10

Hybrid Intrusion Detection with Weighted Signature Generation over Anomalous Internet Episodes Kai Hwang, Fellow, Hybrid Intrusion Detection with Weighted Signature Generation over Anomalous Internet Episodes Kai Hwang, Fellow, IEEE, Min Cai, Member, IEEE, Ying Chen, Student Member, IEEE, and Min Qin IEEE TRANSACTIONS ON DEPENDABLE AND SECURE COMPUTING, VOL. 4, NO. 1, JANUARY-MARCH 2007 Presented by Yong Sun Kim

Summary This hybrid system combines ¾ advantage of low false-positive rate of signature -based Summary This hybrid system combines ¾ advantage of low false-positive rate of signature -based IDS(intrusion detection system) ¾ ability of detect novel unknown attacks of ADS(anomaly detection system) Experimental results show ¾ 60 percent detection rate of the HIDS, compared with 30 percent and 22 percent in using the SNORT and Bro systems, respectively. And it was obtained with less than 3 percent false alarms. ¾ The signatures generated by ADS upgrade the SNORT performance by 33 percent.

A hybrid intrusion detection system built with a SNORT and an anomaly detection subsystem A hybrid intrusion detection system built with a SNORT and an anomaly detection subsystem (ADS) through automated signature generation from Internet episodes.

Data mining scheme for network anomaly detection over Internet connection records. The anomaly is Data mining scheme for network anomaly detection over Internet connection records. The anomaly is detected once the episode rule cannot find any match with normal connection rules in database. Use the attack data set as a mixture of locally captured trace file and DARPA 1999 IDS evaluation data set(MIT/LL).

Appreciative Comment 1 To maximize the effectiveness, various algorithms and schemes are introduced ¾ Appreciative Comment 1 To maximize the effectiveness, various algorithms and schemes are introduced ¾ Mining FERs( frequent episode rules) for anomaly detection ¾ Episode Rule Training from Normal Traffic ¾ Pruning techniques for episode rules

Appreciative Comment 2 By using a weighted signature generation algorithm, improve accuracy and reduce Appreciative Comment 2 By using a weighted signature generation algorithm, improve accuracy and reduce false alarms ADS assigns an anomaly score and a normality score for each connection after processing a traffic data set ¾ Define signatures when patterns have high anomaly scores but relatively low normality scores. ¾ The

Critical Comments There exist a different description about False-Alarm Rate and Detection Rate ¾ Critical Comments There exist a different description about False-Alarm Rate and Detection Rate ¾ ¾ ¾ In Abstract : ”results show a 60 percent… less than 3 percent false alarm. . ” In 8 Conclusions and Further Research 4. : ”Our HIDS results in a detection rate of 60 percent… false alarms must be maintained below 3 percent. ” In 7. 3 Effects of False Alarms on IDS Performance : ”The HIDS achieved a low 47 percent detection rate at 1 percent false alarms. However, the detection rate can be raised to 60 percent if the false alarms can be tolerated up to 30 percent”

Fig. 13. ROC curves showing the variation of the average intrusion detection rate of Fig. 13. ROC curves showing the variation of the average intrusion detection rate of three detection systems as the false alarm rate increases.

Question Is this passive way which generates signature in ADS still effective under the Question Is this passive way which generates signature in ADS still effective under the fast network attack such as “Code Red” ?

Signature mapping Dataset-I, the < attribute; condition > pair is decoded as follows: (ip Signature mapping Dataset-I, the < attribute; condition > pair is decoded as follows: (ip proto = icmp), (icmp type = echo req), (1, 480 <= src bytes < 1, 490), (dst count > 10) The < attribute; condition > pairs form an abstract signature of the Pod attack. Using the attribute mappings in Table 4, we translate the signature into a SNORT rule as follows: alert icmp$EXTERNAL NET any <> $HOME NET any (msg : ”possible pod attack”; itype : 8; dsize : 1, 480 <> 1, 490; threshold : type both, track by_dst, count 10 seconds 1; sid : 900, 001; rev : 0; ).