29
Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against worms o Detect the worm early o Slow-down and impair worm propagation Summary from the Last Lecture

Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

  • View
    218

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Looked at some research approaches to:o Evaluate defense effectivenesso Stop worm from spreading from a given hosto Defend a circle of friends against wormso Detect the worm earlyo Slow-down and impair worm propagation

Summary from the Last Lecture

Page 2: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

The goal is to build an overlay network so that nodes cooperatively detect intrusion activityoCooperation reduces the number of false

positivesOverlay can be used for worm detectionMain feature are active-sink nodes that

detect traffic to unused IP addressesThe reaction is to build blacklists of

infected nodes

DOMINO

V. Yegneswaran, P. Barford, S. Jha, “Global Intrusion Detection in the DOMINOOverlay System,” NDSS 2004

Page 3: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

DOMINO Architecture

Page 4: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Axis nodes collect, aggregate and share dataoNodes in large, trustworthy ISPsoEach node maintains a NIDS and an active sink

over large portion of unused IP spaceAccess points grant access to axis nodes

after thorough administrative checksSatellite nodes form trees below an axis

node, collect information and deliver it to axis nodes and pull relevant information

Terrestrial nodes supply daily summaries of port scan data

DOMINO Architecture

Page 5: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Every axis node maintains a global and local view of intrusion activity

Periodically a node receives summaries from peers which are used to update global viewoList of worst offenders grouped per portoLists of top scanned ports

RSA is used to authenticate nodes and signed SHA digests are used to ensure message integrity and authenticity

Information Sharing

Page 6: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

How Many Nodes We Need?

40 for port summaries

20 for worst offender list

Page 7: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

How Frequent Info Exchange?Staleness doesn’t matter much but more frequent lists are better to catch worst offenders

Page 8: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

How Long Blacklists?About 1000 IPs are enough

Page 9: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

How Close Monitoring Nodes?Blacklists in same /16 space are similar satellites in /16 space should be groupedunder the same axis node and sets of /16 spaces should be randomly distributed

among different axis nodes

Page 10: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Slow worm propagated in May 2002Nodes exchange reports hourlyAlarm is raised if 20% or more nodes vote

for an alarmA node votes if all of these hold:o200% increase in number of scans from hourly

averageo100% increase in sources from hourly averageoNumber of sources > 5

SQL Snake Experiments

Page 11: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

SQL Snake Reaction Time

Almost zero

Page 12: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Extremely fast worm periodic information exchange will not be enough

We need spontaneous alerts issued through triggers

A trigger is issued if it holds:oNumber of sources > 5, andoRule 1: Number of scans is 10 times the average, oroRule 2: Number of sources is 10 times the average,

oroRule 3: The duration of anomalous event (horizontal,

vertical or coordinated scan) is 10 times the average Detection is called if more than 10% (Rule1),

20% (Rule 2) or 30% (Rule 3) nodes issue alerts

SQL Slammer Experiments

Page 13: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

SQL Slammer Reaction Time

About 80-100 class C subnets are enough

Page 14: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Focus on TCP worms that propagate via scanning

Idea: vulnerability exploit is not easily mutable so worm packets should have some common signature

Step 1: Select suspicious TCP flows using heuristics

Step 2: Generate signatures using content prevalence analysis

Automatic Worm Signatures

Kim, H.-A. and Karp, B., Autograph: Toward Automated, Distributed Worm Signature Detection, in the Proceedings of the 13th Usenix Security

Symposium (Security 2004), San Diego, CA, August, 2004.

Page 15: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Detect scanners as hosts that make many unsuccessful connection attempts (>2)

Select their successful flows as suspiciousBuild suspicious flow pooloWhen there’s enough flows inside trigger

signature generation step

Suspicious Flows

Page 16: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Use most frequent byte sequences across flows as the signature

Naïve techniques fail at byte insertion, deletion, reordering

Content-based payload partitioning (COPP)o Partition if Rabin fingerprint of a sliding window

matches breakmark = content blockso Configurable parameters: window size,

breakmarko Analyze which content blocks appear most

frequently and what is the smallest set of those that covers most/all samples in suspicious flow pool

Signature Generation

Page 17: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Tested on traces of HTTP traffic interlaced with known worms

For large block sizes and large coverage of suspicious flow pool (90-95%) Autograph performs very wello Small false positives and false negatives

How Well Does it Work?

Page 18: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Would detect more scannersWould produce more data for suspicious

flow poolo Reduce false positives and false negatives

Distributed Autograph

Page 19: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Detect content prevalence o Some content may vary but some portion of

worm remains invariantDetect address dispersiono Same content will be sent from many hosts to

many destinationsChallenge: how to detect these efficiently

(low cost = fast operation)

Automatic Signatures (approach 2)

S.Singh, C. Estan, G. Varghese and S. Savage “Automated Worm Fingerprinting,” OSDI 2004

Page 20: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Hash content + port + proto and use this as key to a table where counters are kepto Content hash is calculated over overlapping

blocks of fixed size o Use Rabin fingerprint as hash functiono Autograph calculates Rabin fingerprint over

variable-length blocks that are non-overlapping

Content Prevalence Detection

Page 21: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Remembering sources and destinations for each content would require too much memory

Scaled bitmap:o Sample down input space, e.g., hash into values

0-63 but only remember those values that hash into 0-31

o Set the bit for the output value (out of 32 bits)o Increase sampling-down factor each time bitmap

is full = constant space, flexible counting

Address Dispersion Detection

Page 22: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Implemented and deployed at UCSD network

How Well Does This Work?

Page 23: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Some false positiveso Spam, common HTTP protocol headers .. (easily

whitelisted)o Popular BitTorrent files (not easily whitelisted)

No false negativeso Detected each worm outbreak reported in newso Cross-checked with Snort’s signature detection

How Well Does This Work?

Page 24: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Insight: multiple invariant substrings must be present in all variants of the worm for the exploit to worko Protocol framing (force the vulnerable code

down the path where the vulnerability exists)o Return address

Substrings not enough = too shortSignature: multiple disjoint byte stringso Conjunction of byte stringso Token subsequences (must appear in order)o Bayes-scored substrings (score + threshold)

Polymorphic Worm Signatures

J. Newsome, B. Karp and D. Song, “Polygraph: Automatically Generating Signatures for Polymorphic Worms,”

IEEE Security and Privacy Symposium, 2005

Page 25: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Invariant bytes: any change makes the worm fail

Wildcard bytes: any change has no effectCode bytes: Can be changed using some

polymorphic technique and worm will still worko E.g., encryption

Worm Code Structure

Page 26: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

All traffic is seen, some is identified as part of suspicious flows and sent to suspicious traffic pooloMay contain some good trafficoMay contain multiple worms

Rest of traffic is sent to good traffic poolAlgorithm makes a single pass over pools

and generates signatures

Polygraph Architecture

Page 27: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Extract tokens (variable length) that occur in at least K sampleso Conjuction signature is this set of tokenso To find token-subsequence signatures samples in

the pool are aligned in different ways (shifted left or right) so that the maximum-length subsequences are identified

o Contiguous tokens are preferredo For Bayes signatures for each token a probability

is computed that it is contained by a good or a suspicious flow – use this as a score

o Set high value of threshold to avoid false positives

Signature Detection

Page 28: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

Legitimate traffic traces: HTTP and DNSo Good traffic poolo Some of this traffic mixed with worm traffic to

model imperfect separationWorm traffic: Ideally-polymorphic worms

generated from 3 known exploitsVarious tests conducted

How Well Does This Work?

Page 29: Looked at some research approaches to: o Evaluate defense effectiveness o Stop worm from spreading from a given host o Defend a circle of friends against

When compared with single signature (longest substring) detection, all proposed signatures result in lower false positive rateso False negative rate is always zero if the

suspicious pool has at least three samples If some good traffic ends up in suspicious

poolo False negative rate is still lowo False positive rate is low until noise gets too big

If there are multiple worms in suspicious pool and noiseo False positives and false negatives are still low

How Well Does This Work?