def find_lib_cluster_link(so_inst, libs):
(closest_relative, max_similarity) = ("nothing", 0.0)
for lib in libs:
# lib's with absent sigs cannot be cluster centers
if lib["jni_onload_info"]["signature"] != []:
print "Comparing signatures of " + lib["so_file_name"] + " and " + so_inst.so_file_name
print lib["jni_onload_info"]["signature"], so_inst.mnemonics
# calculate weight distance based on the shorter lengthed signature
# TODO: Remove similar code at the start of JNI_OnLoad's
(distance, num_mnemonics) = tapered_levenshtein(lib["jni_onload_info"]["signature"], so_inst.mnemonics)
# calc similarity with center
similarity = 1.0 - float(distance) / num_mnemonics
print "\tSimilarity is " + str(similarity) + " and distance " + str(distance) + "\n"
if similarity > max_similarity:
# found cluster, end search
# TODO: switch to insert_many with large amounts of apks
max_similarity = similarity
closest_relative = lib["so_file_name"]
print "Chose " + closest_relative
return (closest_relative, max_similarity)
def main(options, arguments):
libs = libraries.find()
for json_obj in libs:
for comp_obj in libs:
(distance, num_mnemonics) = tapered_levenshtein(center_obj["jni_onload_info"]["signature"], so_inst.mnemonics)
if distance != -1.0:
print 'no mnemonics'
break
# calc similarity with center
similarity = 1.0 - float(distance)/num_mnemonics
if similarity >= 0.9:
# found cluster, end search
# TODO: switch to insert_many with large amounts of apks
instance = libraries.insert_one({'so_file_name': so_inst.so_file_name, 'arch': so_inst.arch, 'apks_found_in':[so_inst.apk_filename], 'hash': so_inst.sha1, 'jni_onload_info': {'signature': so_inst.mnemonics,'is_cluster_center': False, 'similarity_with_center': similarity, 'variations': [] } })
print 'added to existing cluster'
return
